Skip to content

Sentiment Analysis

Aspect-based sentiment analysis predictive task.

FewshotExample

Bases: FewshotExample

Few-shot example with per-aspect sentiment scores.

Source code in sieves/tasks/predictive/sentiment_analysis/core.py 
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
class FewshotExample(BaseFewshotExample):
    """Few-shot example with per-aspect sentiment scores."""

    sentiment_per_aspect: dict[str, float]

    @override
    @property
    def target_fields(self) -> Sequence[str]:
        return ("sentiment_per_aspect",)

    @pydantic.model_validator(mode="after")
    def check_confidence(self) -> FewshotExample:
        """Validate that 'overall' exists and all scores are in [0, 1]."""
        assert "overall" in self.sentiment_per_aspect, ValueError(
            "'overall' score has to be given in `sentiment_per_aspect` dict."
        )
        if any([conf for conf in self.sentiment_per_aspect.values() if not 0 <= conf <= 1]):
            raise ValueError("Sentiment score has to be between 0 and 1.")
        return self

input_fields property

Defines which fields are inputs.

Returns:

Type Description
Sequence[str]

Sequence of field names.

check_confidence()

Validate that 'overall' exists and all scores are in [0, 1].

Source code in sieves/tasks/predictive/sentiment_analysis/core.py 
42
43
44
45
46
47
48
49
50
@pydantic.model_validator(mode="after")
def check_confidence(self) -> FewshotExample:
    """Validate that 'overall' exists and all scores are in [0, 1]."""
    assert "overall" in self.sentiment_per_aspect, ValueError(
        "'overall' score has to be given in `sentiment_per_aspect` dict."
    )
    if any([conf for conf in self.sentiment_per_aspect.values() if not 0 <= conf <= 1]):
        raise ValueError("Sentiment score has to be between 0 and 1.")
    return self

from_dspy(example) classmethod

Convert from dspy.Example.

Parameters:

Name Type Description Default
example Example

Example as dspy.Example.

 required

Returns:

Type Description
Self

Example as FewshotExample.
Source code in sieves/tasks/predictive/core.py 
74
75
76
77
78
79
80
81
@classmethod
def from_dspy(cls, example: dspy.Example) -> Self:
    """Convert from `dspy.Example`.

    :param example: Example as `dspy.Example`.
    :returns: Example as `FewshotExample`.
    """
    return cls(**example)

to_dspy()

Convert to dspy.Example.

Returns:

Type Description
Example

Example as dspy.Example.
Source code in sieves/tasks/predictive/core.py 
67
68
69
70
71
72
def to_dspy(self) -> dspy.Example:
    """Convert to `dspy.Example`.

    :returns: Example as `dspy.Example`.
    """
    return dspy.Example(**Engine.convert_fewshot_examples([self])[0]).with_inputs(*self.input_fields)

SentimentAnalysis

Bases: PredictiveTask[_TaskPromptSignature, _TaskResult, _TaskBridge]

Estimate per‑aspect and overall sentiment for a document.

Source code in sieves/tasks/predictive/sentiment_analysis/core.py 
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
class SentimentAnalysis(PredictiveTask[_TaskPromptSignature, _TaskResult, _TaskBridge]):
    """Estimate per‑aspect and overall sentiment for a document."""

    def __init__(
        self,
        model: _TaskModel,
        generation_settings: GenerationSettings = GenerationSettings(),
        aspects: tuple[str, ...] = tuple(),
        task_id: str | None = None,
        include_meta: bool = True,
        batch_size: int = -1,
        prompt_instructions: str | None = None,
        fewshot_examples: Sequence[FewshotExample] = (),
        condition: Callable[[Doc], bool] | None = None,
    ) -> None:
        """
        Initialize SentimentAnalysis task.

        :param model: Model to use.
        :param generation_settings: Settings for structured generation.
        :param aspects: Aspects to consider in sentiment analysis. Overall sentiment will always be determined. If
            empty, only overall sentiment will be determined.
        :param task_id: Task ID.
        :param include_meta: Whether to include meta information generated by the task.
        :param batch_size: Batch size to use for inference. Use -1 to process all documents at once.
        :param prompt_instructions: Custom prompt instructions. If None, default instructions are used.
        :param fewshot_examples: Few-shot examples.
        :param condition: Optional callable that determines whether to process each document.
        """
        self._aspects = tuple(sorted(set(aspects) | {"overall"}))
        super().__init__(
            model=model,
            generation_settings=generation_settings,
            task_id=task_id,
            include_meta=include_meta,
            batch_size=batch_size,
            overwrite=False,
            prompt_instructions=prompt_instructions,
            fewshot_examples=fewshot_examples,
            condition=condition,
        )
        self._fewshot_examples: Sequence[FewshotExample]

    @override
    def _init_bridge(self, engine_type: EngineType) -> _TaskBridge:
        bridge_types: dict[EngineType, type[_TaskBridge]] = {
            EngineType.dspy: DSPySentimentAnalysis,
            EngineType.outlines: OutlinesSentimentAnalysis,
            EngineType.langchain: LangChainSentimentAnalysis,
        }

        try:
            bridge_type = bridge_types[engine_type]

            return bridge_type(
                task_id=self._task_id,
                prompt_instructions=self._custom_prompt_instructions,
                aspects=self._aspects,
                generation_settings=self._generation_settings,
            )
        except KeyError as err:
            raise KeyError(f"Engine type {engine_type} is not supported by {self.__class__.__name__}.") from err

    @staticmethod
    @override
    def supports() -> set[EngineType]:
        return {
            EngineType.dspy,
            EngineType.langchain,
            EngineType.outlines,
        }

    @override
    def _validate_fewshot_examples(self) -> None:
        for fs_example in self._fewshot_examples or []:
            if any([aspect not in self._aspects for aspect in fs_example.sentiment_per_aspect]) or not all(
                [label in fs_example.sentiment_per_aspect for label in self._aspects]
            ):
                raise ValueError(
                    f"Aspect mismatch: {self._task_id} has aspects {self._aspects}. Few-shot examples have "
                    f"aspects {fs_example.sentiment_per_aspect.keys()}."
                )

    @property
    @override
    def _state(self) -> dict[str, Any]:
        return {
            **super()._state,
            "aspects": self._aspects,
        }

    @override
    def to_hf_dataset(self, docs: Iterable[Doc], threshold: float = 0.5) -> datasets.Dataset:
        # Define metadata.
        features = datasets.Features(
            {"text": datasets.Value("string"), "aspect": datasets.Sequence(datasets.Value("float32"))}
        )
        info = datasets.DatasetInfo(
            description=f"Aspect-based sentiment analysis dataset with aspects {self._aspects}. Generated with sieves "
            f"v{Config.get_version()}.",
            features=features,
        )

        # Fetch data used for generating dataset.
        aspects = self._aspects
        try:
            data = [(doc.text, doc.results[self._task_id]) for doc in docs]
        except KeyError as err:
            raise KeyError(f"Not all documents have results for this task with ID {self._task_id}") from err

        def generate_data() -> Iterable[dict[str, Any]]:
            """Yield results as dicts.

            :return: Results as dicts.
            """
            for text, result in data:
                scores = {sent_score[0]: sent_score[1] for sent_score in result}
                yield {"text": text, "aspect": [scores[aspect] for aspect in aspects]}

        # Create dataset.
        return datasets.Dataset.from_generator(generate_data, features=features, info=info)

    @override
    def distill(
        self,
        base_model_id: str,
        framework: DistillationFramework,
        data: datasets.Dataset | Sequence[Doc],
        output_path: Path | str,
        val_frac: float,
        init_kwargs: dict[str, Any] | None = None,
        train_kwargs: dict[str, Any] | None = None,
        seed: int | None = None,
    ) -> None:
        raise NotImplementedError

    @override
    def _evaluate_optimization_example(
        self, truth: dspy.Example, pred: dspy.Prediction, trace: Any, model: dspy.LM
    ) -> float:
        # Compute per-aspect accuracy as 1 - abs(true_sentiment - pred_sentiment)
        # Average across all aspects (same approach as multi-label classification)
        accuracy = 0
        for aspect, sentiment in truth["sentiment_per_aspect"].items():
            if aspect in pred["sentiment_per_aspect"]:
                pred_sentiment = max(min(pred["sentiment_per_aspect"][aspect], 1), 0)
                accuracy += 1 - abs(sentiment - pred_sentiment)

        return accuracy / len(truth["sentiment_per_aspect"])

fewshot_examples property

Return few-shot examples.

Returns:

Type Description
Sequence[FewshotExample]

Few-shot examples.

id property

Return task ID.

Used by pipeline for results and dependency management.

Returns:

Type Description
str

Task ID.

prompt_signature_description property

Return prompt signature description.

Returns:

Type Description
str | None

Prompt signature description.

prompt_template property

Return prompt template.

Returns:

Type Description
str

Prompt template.

__add__(other)

Chain this task with another task or pipeline using the + operator.

This returns a new Pipeline that executes this task first, followed by the task(s) in other. The original task(s)/pipeline are not mutated.

Cache semantics: - If other is a Pipeline, the resulting pipeline adopts other's use_cache setting (because the left-hand side is a single task). - If other is a Task, the resulting pipeline defaults to use_cache=True.

Parameters:

Name Type Description Default
other Task | Pipeline

A Task or Pipeline to execute after this task.

 required

Returns:

Type Description
Pipeline

A new Pipeline representing the chained execution.

Raises:

Type Description
TypeError

If other is not a Task or Pipeline.
Source code in sieves/tasks/core.py 
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
def __add__(self, other: Task | Pipeline) -> Pipeline:
    """Chain this task with another task or pipeline using the ``+`` operator.

    This returns a new ``Pipeline`` that executes this task first, followed by the
    task(s) in ``other``. The original task(s)/pipeline are not mutated.

    Cache semantics:
    - If ``other`` is a ``Pipeline``, the resulting pipeline adopts ``other``'s
      ``use_cache`` setting (because the left-hand side is a single task).
    - If ``other`` is a ``Task``, the resulting pipeline defaults to ``use_cache=True``.

    :param other: A ``Task`` or ``Pipeline`` to execute after this task.
    :return: A new ``Pipeline`` representing the chained execution.
    :raises TypeError: If ``other`` is not a ``Task`` or ``Pipeline``.
    """
    # Lazy import to avoid circular dependency at module import time.
    from sieves.pipeline import Pipeline

    if isinstance(other, Pipeline):
        return Pipeline(tasks=[self, *other.tasks], use_cache=other.use_cache)

    if isinstance(other, Task):
        return Pipeline(tasks=[self, other])

    raise TypeError(f"Cannot chain Task with {type(other).__name__}")

__call__(docs)

Execute task with conditional logic.

Checks the condition for each document without materializing all docs upfront. Passes all documents that pass the condition to _call() for proper batching. Documents that fail the condition have results[task_id] set to None.

Parameters:

Name Type Description Default
docs Iterable[Doc]

Docs to process.

 required

Returns:

Type Description
Iterable[Doc]

Processed docs (in original order).
Source code in sieves/tasks/core.py 
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
def __call__(self, docs: Iterable[Doc]) -> Iterable[Doc]:
    """Execute task with conditional logic.

    Checks the condition for each document without materializing all docs upfront.
    Passes all documents that pass the condition to _call() for proper batching.
    Documents that fail the condition have results[task_id] set to None.

    :param docs: Docs to process.
    :return: Processed docs (in original order).
    """
    # Create three independent iterators:
    #   1. Check which docs pass condition.
    #   2. Yield only passing docs to _call().
    #   3. Iterate and yield results in order.
    docs_iters = itertools.tee(docs, 3)

    # First pass: determine which docs pass the condition by index
    passing_indices: set[int] = set()

    for idx, doc in enumerate(docs_iters[0]):
        if self._condition is None or self._condition(doc):
            passing_indices.add(idx)

    # Process all passing docs together.
    processed = self._call(d for i, d in enumerate(docs_iters[1]) if i in passing_indices)
    processed_iter = iter(processed) if not isinstance(processed, Iterator) else processed

    # Iterate through original docs in order and yield results
    for idx, doc in enumerate(docs_iters[2]):
        if idx in passing_indices:
            # Doc passed condition - use processed result.
            yield next(processed_iter)
        else:
            # Doc failed condition - set None result and yield original.
            doc.results[self.id] = None
            yield doc

__init__(model, generation_settings=GenerationSettings(), aspects=tuple(), task_id=None, include_meta=True, batch_size=-1, prompt_instructions=None, fewshot_examples=(), condition=None)

Initialize SentimentAnalysis task.

Parameters:

Name Type Description Default
model _TaskModel

Model to use.

 required
generation_settings GenerationSettings

Settings for structured generation.

 GenerationSettings()
aspects tuple[str, ...]

Aspects to consider in sentiment analysis. Overall sentiment will always be determined. If empty, only overall sentiment will be determined.

 tuple()
task_id str | None

Task ID.

 None
include_meta bool

Whether to include meta information generated by the task.

 True
batch_size int

Batch size to use for inference. Use -1 to process all documents at once.

 -1
prompt_instructions str | None

Custom prompt instructions. If None, default instructions are used.

 None
fewshot_examples Sequence[FewshotExample]

Few-shot examples.

 ()
condition Callable[[Doc], bool] | None

Optional callable that determines whether to process each document.

 None
Source code in sieves/tasks/predictive/sentiment_analysis/core.py 
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
def __init__(
    self,
    model: _TaskModel,
    generation_settings: GenerationSettings = GenerationSettings(),
    aspects: tuple[str, ...] = tuple(),
    task_id: str | None = None,
    include_meta: bool = True,
    batch_size: int = -1,
    prompt_instructions: str | None = None,
    fewshot_examples: Sequence[FewshotExample] = (),
    condition: Callable[[Doc], bool] | None = None,
) -> None:
    """
    Initialize SentimentAnalysis task.

    :param model: Model to use.
    :param generation_settings: Settings for structured generation.
    :param aspects: Aspects to consider in sentiment analysis. Overall sentiment will always be determined. If
        empty, only overall sentiment will be determined.
    :param task_id: Task ID.
    :param include_meta: Whether to include meta information generated by the task.
    :param batch_size: Batch size to use for inference. Use -1 to process all documents at once.
    :param prompt_instructions: Custom prompt instructions. If None, default instructions are used.
    :param fewshot_examples: Few-shot examples.
    :param condition: Optional callable that determines whether to process each document.
    """
    self._aspects = tuple(sorted(set(aspects) | {"overall"}))
    super().__init__(
        model=model,
        generation_settings=generation_settings,
        task_id=task_id,
        include_meta=include_meta,
        batch_size=batch_size,
        overwrite=False,
        prompt_instructions=prompt_instructions,
        fewshot_examples=fewshot_examples,
        condition=condition,
    )
    self._fewshot_examples: Sequence[FewshotExample]

deserialize(config, **kwargs) classmethod

Generate PredictiveTask instance from config.

Parameters:

Name Type Description Default
config Config

Config to generate instance from.

 required
kwargs dict[str, Any]

Values to inject into loaded config.

 {}

Returns:

Type Description
PredictiveTask[TaskPromptSignature, TaskResult, TaskBridge]

Deserialized PredictiveTask instance.
Source code in sieves/tasks/predictive/core.py 
239
240
241
242
243
244
245
246
247
248
249
250
251
252
@classmethod
def deserialize(
    cls, config: Config, **kwargs: dict[str, Any]
) -> PredictiveTask[TaskPromptSignature, TaskResult, TaskBridge]:
    """Generate PredictiveTask instance from config.

    :param config: Config to generate instance from.
    :param kwargs: Values to inject into loaded config.
    :return PredictiveTask[_TaskPromptSignature, _TaskResult, _TaskBridge]: Deserialized PredictiveTask instance.
    """
    init_dict = config.to_init_dict(cls, **kwargs)
    init_dict["generation_settings"] = GenerationSettings.model_validate(init_dict["generation_settings"])

    return cls(**init_dict)

optimize(optimizer, verbose=True)

Optimize task prompt and few-shot examples with the available optimization config.

Updates task to use best prompt and few-shot examples found by the optimizer.

Parameters:

Name Type Description Default
optimizer Optimizer

Optimizer to run.

 required
verbose bool

Whether to suppress output. DSPy produces a good amount of logs, so this can be useful to not pollute your terminal. Only warnings and errors will be printed.

 True

Returns:

Type Description
tuple[str, Sequence[FewshotExample]]

Best found prompt and few-shot examples.
Source code in sieves/tasks/predictive/core.py 
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
def optimize(self, optimizer: optimization.Optimizer, verbose: bool = True) -> tuple[str, Sequence[FewshotExample]]:
    """Optimize task prompt and few-shot examples with the available optimization config.

    Updates task to use best prompt and few-shot examples found by the optimizer.

    :param optimizer: Optimizer to run.
    :param verbose: Whether to suppress output. DSPy produces a good amount of logs, so this can be useful to
        not pollute your terminal. Only warnings and errors will be printed.

    :return tuple[str, Sequence[FewshotExample]]: Best found prompt and few-shot examples.
    """
    assert len(self._fewshot_examples) > 1, "At least two few-shot examples need to be provided to optimize."

    # Run optimizer to get best prompt and few-shot examples.
    signature = self._get_task_signature()
    dspy_examples = [ex.to_dspy() for ex in self._fewshot_examples]

    def _pred_eval(truth: dspy.Example, pred: dspy.Prediction, trace: Any | None = None) -> float:
        """Wraps optimization evaluation, injects model.

        :param truth: Ground truth.
        :param pred: Predicted value.
        :param trace: Optional trace information.
        :return: Metric value between 0.0 and 1.0.
        :raises KeyError: If target fields are missing from truth or prediction.
        :raises ValueError: If similarity score cannot be parsed from LLM response.
        """
        return self._evaluate_optimization_example(truth, pred, trace, model=optimizer.model)

    if verbose:
        best_prompt, best_examples = optimizer(signature, dspy_examples, _pred_eval, verbose=verbose)
    else:
        # Temporarily suppress DSPy logs.
        dspy_logger = logging.getLogger("dspy")
        optuna_logger = logging.getLogger("optuna")
        original_dspy_level = dspy_logger.level
        original_optuna_level = optuna_logger.level

        try:
            dspy_logger.setLevel(logging.ERROR)
            optuna_logger.setLevel(logging.ERROR)
            with warnings.catch_warnings():
                warnings.simplefilter("ignore")
                best_prompt, best_examples = optimizer(signature, dspy_examples, _pred_eval, verbose=verbose)
        finally:
            dspy_logger.setLevel(original_dspy_level)
            optuna_logger.setLevel(original_optuna_level)

    # Update few-shot examples and prompt instructions.
    fewshot_example_cls = self._fewshot_examples[0].__class__
    self._fewshot_examples = [fewshot_example_cls.from_dspy(ex) for ex in best_examples]
    self._validate_fewshot_examples()
    self._custom_prompt_instructions = best_prompt

    # Reinitialize bridge to use new prompt and few-shot examples.
    self._bridge = self._init_bridge(EngineType.get_engine_type(self._engine))

    return best_prompt, self._fewshot_examples

serialize()

Serialize task.

Returns:

Type Description
Config

Config instance.
Source code in sieves/tasks/core.py 
137
138
139
140
141
142
def serialize(self) -> Config:
    """Serialize task.

    :return: Config instance.
    """
    return Config.create(self.__class__, {k: Attribute(value=v) for k, v in self._state.items()})

Bridges for sentiment analysis task.

DSPySentimentAnalysis

Bases: SentAnalysisBridge[PromptSignature, Result, InferenceMode]

DSPy bridge for sentiment analysis.

Source code in sieves/tasks/predictive/sentiment_analysis/bridges.py 
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
class DSPySentimentAnalysis(SentAnalysisBridge[dspy_.PromptSignature, dspy_.Result, dspy_.InferenceMode]):
    """DSPy bridge for sentiment analysis."""

    @override
    @property
    def _default_prompt_instructions(self) -> str:
        return """
        Aspect-based sentiment analysis of the provided text given the provided aspects.
        For each aspect, provide the sentiment score with which you reflects the sentiment in the provided text with
        respect to this aspect.
        The "overall" aspect should reflect the sentiment in the text overall.
        A score of 1.0 means that the sentiment in the text with respect to this aspect is extremely positive.
        0 means the opposite, 0.5 means neutral.
        Sentiment per aspect should always be between 0 and 1.
        """

    @override
    @property
    def _prompt_example_template(self) -> str | None:
        return None

    @override
    @property
    def _prompt_conclusion(self) -> str | None:
        return None

    @override
    @cached_property
    def prompt_signature(self) -> type[dspy_.PromptSignature]:
        aspects = self._aspects
        # Dynamically create Literal as output type.
        AspectType = Literal[*aspects]  # type: ignore[valid-type]

        class SentimentAnalysis(dspy.Signature):  # type: ignore[misc]
            text: str = dspy.InputField(description="Text to determine sentiments for.")
            sentiment_per_aspect: dict[AspectType, float] = dspy.OutputField(
                description="Sentiment in this text with respect to the corresponding aspect."
            )

        SentimentAnalysis.__doc__ = jinja2.Template(self._prompt_instructions).render()

        return SentimentAnalysis

    @override
    @property
    def inference_mode(self) -> dspy_.InferenceMode:
        return self._generation_settings.inference_mode or dspy_.InferenceMode.predict

    @override
    def integrate(self, results: Iterable[dspy_.Result], docs: Iterable[Doc]) -> Iterable[Doc]:
        for doc, result in zip(docs, results):
            assert len(result.completions.sentiment_per_aspect) == 1
            sorted_preds = sorted(
                ((aspect, score) for aspect, score in result.completions.sentiment_per_aspect[0].items()),
                key=lambda x: x[1],
                reverse=True,
            )
            doc.results[self._task_id] = sorted_preds
        return docs

    @override
    def consolidate(
        self, results: Iterable[dspy_.Result], docs_offsets: list[tuple[int, int]]
    ) -> Iterable[dspy_.Result]:
        results = list(results)

        # Determine label scores for chunks per document.
        for doc_offset in docs_offsets:
            aspect_scores: dict[str, float] = {label: 0.0 for label in self._aspects}
            doc_results = results[doc_offset[0] : doc_offset[1]]

            for res in doc_results:
                if res is None:
                    continue

                assert len(res.completions.sentiment_per_aspect) == 1
                for label, score in res.completions.sentiment_per_aspect[0].items():
                    # Clamp score to range between 0 and 1. Alternatively we could force this in the prompt signature,
                    # but this fails occasionally with some models and feels too strict (maybe a strict mode would be
                    # useful?).
                    aspect_scores[label] += max(0, min(score, 1))

            sorted_aspect_scores: list[dict[str, str | float]] = sorted(
                (
                    {"aspect": aspect, "score": score / (doc_offset[1] - doc_offset[0])}
                    for aspect, score in aspect_scores.items()
                ),
                key=lambda x: x["score"],
                reverse=True,
            )

            yield dspy.Prediction.from_completions(
                {
                    "sentiment_per_aspect": [{sls["aspect"]: sls["score"] for sls in sorted_aspect_scores}],
                },
                signature=self.prompt_signature,
            )

prompt_template property

Return prompt template.

Chains _prompt_instructions, _prompt_example_template and _prompt_conclusion.

Note: different engines have different expectations as how a prompt should look like. E.g. outlines supports the Jinja 2 templating format for insertion of values and few-shot examples, whereas DSPy integrates these things in a different value in the workflow and hence expects the prompt not to include these things. Mind engine-specific expectations when creating a prompt template.

Returns:

Type Description
str

Prompt template as string. None if not used by engine.

__init__(task_id, prompt_instructions, aspects, generation_settings)

Initialize SentAnalysisBridge.

Parameters:

Name Type Description Default
task_id str

Task ID.

 required
prompt_instructions str | None

Custom prompt instructions. If None, default instructions are used.

 required
aspects tuple[str, ...]

Aspects to consider.

 required
generation_settings GenerationSettings

Generation settings including inference_mode.

 required
Source code in sieves/tasks/predictive/sentiment_analysis/bridges.py 
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
def __init__(
    self,
    task_id: str,
    prompt_instructions: str | None,
    aspects: tuple[str, ...],
    generation_settings: GenerationSettings,
):
    """Initialize SentAnalysisBridge.

    :param task_id: Task ID.
    :param prompt_instructions: Custom prompt instructions. If None, default instructions are used.
    :param aspects: Aspects to consider.
    :param generation_settings: Generation settings including inference_mode.
    """
    super().__init__(
        task_id=task_id,
        prompt_instructions=prompt_instructions,
        overwrite=False,
        generation_settings=generation_settings,
    )
    self._aspects = aspects

extract(docs)

Extract all values from doc instances that are to be injected into the prompts.

Parameters:

Name Type Description Default
docs Iterable[Doc]

Docs to extract values from.

 required

Returns:

Type Description
Iterable[dict[str, Any]]

All values from doc instances that are to be injected into the prompts
Source code in sieves/tasks/predictive/bridges.py 
118
119
120
121
122
123
124
def extract(self, docs: Iterable[Doc]) -> Iterable[dict[str, Any]]:
    """Extract all values from doc instances that are to be injected into the prompts.

    :param docs: Docs to extract values from.
    :return Iterable[dict[str, Any]]: All values from doc instances that are to be injected into the prompts
    """
    return ({"text": doc.text if doc.text else None} for doc in docs)

LangChainSentimentAnalysis

Bases: PydanticBasedSentAnalysis[InferenceMode]

LangChain bridge for sentiment analysis.

Source code in sieves/tasks/predictive/sentiment_analysis/bridges.py 
273
274
275
276
277
278
279
class LangChainSentimentAnalysis(PydanticBasedSentAnalysis[langchain_.InferenceMode]):
    """LangChain bridge for sentiment analysis."""

    @override
    @property
    def inference_mode(self) -> langchain_.InferenceMode:
        return self._generation_settings.inference_mode or langchain_.InferenceMode.structured

prompt_template property

Return prompt template.

Chains _prompt_instructions, _prompt_example_template and _prompt_conclusion.

Note: different engines have different expectations as how a prompt should look like. E.g. outlines supports the Jinja 2 templating format for insertion of values and few-shot examples, whereas DSPy integrates these things in a different value in the workflow and hence expects the prompt not to include these things. Mind engine-specific expectations when creating a prompt template.

Returns:

Type Description
str

Prompt template as string. None if not used by engine.

__init__(task_id, prompt_instructions, aspects, generation_settings)

Initialize SentAnalysisBridge.

Parameters:

Name Type Description Default
task_id str

Task ID.

 required
prompt_instructions str | None

Custom prompt instructions. If None, default instructions are used.

 required
aspects tuple[str, ...]

Aspects to consider.

 required
generation_settings GenerationSettings

Generation settings including inference_mode.

 required
Source code in sieves/tasks/predictive/sentiment_analysis/bridges.py 
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
def __init__(
    self,
    task_id: str,
    prompt_instructions: str | None,
    aspects: tuple[str, ...],
    generation_settings: GenerationSettings,
):
    """Initialize SentAnalysisBridge.

    :param task_id: Task ID.
    :param prompt_instructions: Custom prompt instructions. If None, default instructions are used.
    :param aspects: Aspects to consider.
    :param generation_settings: Generation settings including inference_mode.
    """
    super().__init__(
        task_id=task_id,
        prompt_instructions=prompt_instructions,
        overwrite=False,
        generation_settings=generation_settings,
    )
    self._aspects = aspects

extract(docs)

Extract all values from doc instances that are to be injected into the prompts.

Parameters:

Name Type Description Default
docs Iterable[Doc]

Docs to extract values from.

 required

Returns:

Type Description
Iterable[dict[str, Any]]

All values from doc instances that are to be injected into the prompts
Source code in sieves/tasks/predictive/bridges.py 
118
119
120
121
122
123
124
def extract(self, docs: Iterable[Doc]) -> Iterable[dict[str, Any]]:
    """Extract all values from doc instances that are to be injected into the prompts.

    :param docs: Docs to extract values from.
    :return Iterable[dict[str, Any]]: All values from doc instances that are to be injected into the prompts
    """
    return ({"text": doc.text if doc.text else None} for doc in docs)

OutlinesSentimentAnalysis

Bases: PydanticBasedSentAnalysis[InferenceMode]

Outlines bridge for sentiment analysis.

Source code in sieves/tasks/predictive/sentiment_analysis/bridges.py 
264
265
266
267
268
269
270
class OutlinesSentimentAnalysis(PydanticBasedSentAnalysis[outlines_.InferenceMode]):
    """Outlines bridge for sentiment analysis."""

    @override
    @property
    def inference_mode(self) -> outlines_.InferenceMode:
        return self._generation_settings.inference_mode or outlines_.InferenceMode.json

prompt_template property

Return prompt template.

Chains _prompt_instructions, _prompt_example_template and _prompt_conclusion.

Note: different engines have different expectations as how a prompt should look like. E.g. outlines supports the Jinja 2 templating format for insertion of values and few-shot examples, whereas DSPy integrates these things in a different value in the workflow and hence expects the prompt not to include these things. Mind engine-specific expectations when creating a prompt template.

Returns:

Type Description
str

Prompt template as string. None if not used by engine.

__init__(task_id, prompt_instructions, aspects, generation_settings)

Initialize SentAnalysisBridge.

Parameters:

Name Type Description Default
task_id str

Task ID.

 required
prompt_instructions str | None

Custom prompt instructions. If None, default instructions are used.

 required
aspects tuple[str, ...]

Aspects to consider.

 required
generation_settings GenerationSettings

Generation settings including inference_mode.

 required
Source code in sieves/tasks/predictive/sentiment_analysis/bridges.py 
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
def __init__(
    self,
    task_id: str,
    prompt_instructions: str | None,
    aspects: tuple[str, ...],
    generation_settings: GenerationSettings,
):
    """Initialize SentAnalysisBridge.

    :param task_id: Task ID.
    :param prompt_instructions: Custom prompt instructions. If None, default instructions are used.
    :param aspects: Aspects to consider.
    :param generation_settings: Generation settings including inference_mode.
    """
    super().__init__(
        task_id=task_id,
        prompt_instructions=prompt_instructions,
        overwrite=False,
        generation_settings=generation_settings,
    )
    self._aspects = aspects

extract(docs)

Extract all values from doc instances that are to be injected into the prompts.

Parameters:

Name Type Description Default
docs Iterable[Doc]

Docs to extract values from.

 required

Returns:

Type Description
Iterable[dict[str, Any]]

All values from doc instances that are to be injected into the prompts
Source code in sieves/tasks/predictive/bridges.py 
118
119
120
121
122
123
124
def extract(self, docs: Iterable[Doc]) -> Iterable[dict[str, Any]]:
    """Extract all values from doc instances that are to be injected into the prompts.

    :param docs: Docs to extract values from.
    :return Iterable[dict[str, Any]]: All values from doc instances that are to be injected into the prompts
    """
    return ({"text": doc.text if doc.text else None} for doc in docs)

PydanticBasedSentAnalysis

Bases: SentAnalysisBridge[BaseModel, BaseModel, EngineInferenceMode], ABC

Base class for Pydantic-based sentiment analysis bridges.

Source code in sieves/tasks/predictive/sentiment_analysis/bridges.py 
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
class PydanticBasedSentAnalysis(
    SentAnalysisBridge[pydantic.BaseModel, pydantic.BaseModel, EngineInferenceMode], abc.ABC
):
    """Base class for Pydantic-based sentiment analysis bridges."""

    @override
    @property
    def _default_prompt_instructions(self) -> str:
        return (
            f"""
        Perform aspect-based sentiment analysis of the provided text given the provided aspects:
        {",".join(self._aspects)}."""
            + """
        For each aspect, provide the sentiment in the provided text with respect to this aspect.
        The "overall" aspect should reflect the sentiment in the text overall.
        A score of 1.0 means that the sentiment in the text with respect to this aspect is extremely positive.
        0 means the opposite, 0.5 means neutral.
        The sentiment score per aspect should ALWAYS be between 0 and 1.

        The output for two aspects ASPECT_1 and ASPECT_2 should look like this:
        <output>
            <aspect_sentiments>
                <aspect_sentiment>
                    <aspect>ASPECT_1</aspect>
                    <sentiment>SENTIMENT_SCORE_1</sentiment>
                <aspect_sentiment>
                <aspect_sentiment>
                    <aspect>ASPECT_2</aspect>
                    <sentiment>SENTIMENT_SCORE_2</sentiment>
                <aspect_sentiment>
            </aspect_sentiments>
        </output>
        """
        )

    @override
    @property
    def _prompt_example_template(self) -> str | None:
        return """
        {% if examples|length > 0 -%}
            <examples>
            {%- for example in examples %}
                <example>
                    <text>{{ example.text }}</text>
                    <output>
                        <aspect_sentiments>
                        {%- for a, s in example.sentiment_per_aspect.items() %}
                            <aspect_sentiment>
                                <aspect>{{ a }}</aspect>
                                <sentiment>{{ s }}</sentiment>
                            </aspect_sentiment>
                        {% endfor -%}
                        </aspect_sentiments>
                    </output>
                </example>
            {% endfor %}
            </examples>
        {% endif -%}
        """

    @override
    @property
    def _prompt_conclusion(self) -> str | None:
        return """
        ========

        <text>{{ text }}</text>
        <output>
        """

    @override
    @cached_property
    def prompt_signature(self) -> type[pydantic.BaseModel]:
        prompt_sig = pydantic.create_model(  # type: ignore[no-matching-overload]
            "SentimentAnalysis",
            __base__=pydantic.BaseModel,
            __doc__="Sentiment analysis of specified text.",
            **{aspect: (float, ...) for aspect in self._aspects},
        )

        assert isinstance(prompt_sig, type) and issubclass(prompt_sig, pydantic.BaseModel)
        return prompt_sig

    @override
    def integrate(self, results: Iterable[pydantic.BaseModel], docs: Iterable[Doc]) -> Iterable[Doc]:
        for doc, result in zip(docs, results):
            label_scores = {k: v for k, v in result.model_dump().items() if k != "reasoning"}
            doc.results[self._task_id] = sorted(
                ((aspect, score) for aspect, score in label_scores.items()), key=lambda x: x[1], reverse=True
            )
        return docs

    @override
    def consolidate(
        self, results: Iterable[pydantic.BaseModel], docs_offsets: list[tuple[int, int]]
    ) -> Iterable[pydantic.BaseModel]:
        results = list(results)

        # Determine label scores for chunks per document.
        for doc_offset in docs_offsets:
            aspect_scores: dict[str, float] = {label: 0.0 for label in self._aspects}
            doc_results = results[doc_offset[0] : doc_offset[1]]

            for rec in doc_results:
                if rec is None:
                    continue  # type: ignore[unreachable]

                for aspect in self._aspects:
                    # Clamp score to range between 0 and 1. Alternatively we could force this in the prompt signature,
                    # but this fails occasionally with some models and feels too strict (maybe a strict mode would be
                    # useful?).
                    aspect_scores[aspect] += max(0, min(getattr(rec, aspect), 1))

            yield self.prompt_signature(
                **{aspect: score / (doc_offset[1] - doc_offset[0]) for aspect, score in aspect_scores.items()},
            )

inference_mode abstractmethod property

Return inference mode.

Returns:

Type Description
EngineInferenceMode

Inference mode.

prompt_template property

Return prompt template.

Chains _prompt_instructions, _prompt_example_template and _prompt_conclusion.

Note: different engines have different expectations as how a prompt should look like. E.g. outlines supports the Jinja 2 templating format for insertion of values and few-shot examples, whereas DSPy integrates these things in a different value in the workflow and hence expects the prompt not to include these things. Mind engine-specific expectations when creating a prompt template.

Returns:

Type Description
str

Prompt template as string. None if not used by engine.

__init__(task_id, prompt_instructions, aspects, generation_settings)

Initialize SentAnalysisBridge.

Parameters:

Name Type Description Default
task_id str

Task ID.

 required
prompt_instructions str | None

Custom prompt instructions. If None, default instructions are used.

 required
aspects tuple[str, ...]

Aspects to consider.

 required
generation_settings GenerationSettings

Generation settings including inference_mode.

 required
Source code in sieves/tasks/predictive/sentiment_analysis/bridges.py 
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
def __init__(
    self,
    task_id: str,
    prompt_instructions: str | None,
    aspects: tuple[str, ...],
    generation_settings: GenerationSettings,
):
    """Initialize SentAnalysisBridge.

    :param task_id: Task ID.
    :param prompt_instructions: Custom prompt instructions. If None, default instructions are used.
    :param aspects: Aspects to consider.
    :param generation_settings: Generation settings including inference_mode.
    """
    super().__init__(
        task_id=task_id,
        prompt_instructions=prompt_instructions,
        overwrite=False,
        generation_settings=generation_settings,
    )
    self._aspects = aspects

extract(docs)

Extract all values from doc instances that are to be injected into the prompts.

Parameters:

Name Type Description Default
docs Iterable[Doc]

Docs to extract values from.

 required

Returns:

Type Description
Iterable[dict[str, Any]]

All values from doc instances that are to be injected into the prompts
Source code in sieves/tasks/predictive/bridges.py 
118
119
120
121
122
123
124
def extract(self, docs: Iterable[Doc]) -> Iterable[dict[str, Any]]:
    """Extract all values from doc instances that are to be injected into the prompts.

    :param docs: Docs to extract values from.
    :return Iterable[dict[str, Any]]: All values from doc instances that are to be injected into the prompts
    """
    return ({"text": doc.text if doc.text else None} for doc in docs)

SentAnalysisBridge

Bases: Bridge[_BridgePromptSignature, _BridgeResult, EngineInferenceMode], ABC

Abstract base class for sentiment analysis bridges.

Source code in sieves/tasks/predictive/sentiment_analysis/bridges.py 
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
class SentAnalysisBridge(Bridge[_BridgePromptSignature, _BridgeResult, EngineInferenceMode], abc.ABC):
    """Abstract base class for sentiment analysis bridges."""

    def __init__(
        self,
        task_id: str,
        prompt_instructions: str | None,
        aspects: tuple[str, ...],
        generation_settings: GenerationSettings,
    ):
        """Initialize SentAnalysisBridge.

        :param task_id: Task ID.
        :param prompt_instructions: Custom prompt instructions. If None, default instructions are used.
        :param aspects: Aspects to consider.
        :param generation_settings: Generation settings including inference_mode.
        """
        super().__init__(
            task_id=task_id,
            prompt_instructions=prompt_instructions,
            overwrite=False,
            generation_settings=generation_settings,
        )
        self._aspects = aspects

inference_mode abstractmethod property

Return inference mode.

Returns:

Type Description
EngineInferenceMode

Inference mode.

prompt_signature abstractmethod property

Create output signature.

E.g.: Signature in DSPy, Pydantic objects in outlines, JSON schema in jsonformers. This is engine-specific.

Returns:

Type Description
type[TaskPromptSignature] | TaskPromptSignature

Output signature object. This can be an instance (e.g. a regex string) or a class (e.g. a Pydantic class).

prompt_template property

Return prompt template.

Chains _prompt_instructions, _prompt_example_template and _prompt_conclusion.

Note: different engines have different expectations as how a prompt should look like. E.g. outlines supports the Jinja 2 templating format for insertion of values and few-shot examples, whereas DSPy integrates these things in a different value in the workflow and hence expects the prompt not to include these things. Mind engine-specific expectations when creating a prompt template.

Returns:

Type Description
str

Prompt template as string. None if not used by engine.

__init__(task_id, prompt_instructions, aspects, generation_settings)

Initialize SentAnalysisBridge.

Parameters:

Name Type Description Default
task_id str

Task ID.

 required
prompt_instructions str | None

Custom prompt instructions. If None, default instructions are used.

 required
aspects tuple[str, ...]

Aspects to consider.

 required
generation_settings GenerationSettings

Generation settings including inference_mode.

 required
Source code in sieves/tasks/predictive/sentiment_analysis/bridges.py 
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
def __init__(
    self,
    task_id: str,
    prompt_instructions: str | None,
    aspects: tuple[str, ...],
    generation_settings: GenerationSettings,
):
    """Initialize SentAnalysisBridge.

    :param task_id: Task ID.
    :param prompt_instructions: Custom prompt instructions. If None, default instructions are used.
    :param aspects: Aspects to consider.
    :param generation_settings: Generation settings including inference_mode.
    """
    super().__init__(
        task_id=task_id,
        prompt_instructions=prompt_instructions,
        overwrite=False,
        generation_settings=generation_settings,
    )
    self._aspects = aspects

consolidate(results, docs_offsets) abstractmethod

Consolidate results for document chunks into document results.

Parameters:

Name Type Description Default
results Iterable[TaskResult]

Results per document chunk.

 required
docs_offsets list[tuple[int, int]]

Chunk offsets per document. Chunks per document can be obtained with results[docs_chunk_offsets[i][0]:docs_chunk_offsets[i][1]].

 required

Returns:

Type Description
Iterable[TaskResult]

Results per document.
Source code in sieves/tasks/predictive/bridges.py 
135
136
137
138
139
140
141
142
143
@abc.abstractmethod
def consolidate(self, results: Iterable[TaskResult], docs_offsets: list[tuple[int, int]]) -> Iterable[TaskResult]:
    """Consolidate results for document chunks into document results.

    :param results: Results per document chunk.
    :param docs_offsets: Chunk offsets per document. Chunks per document can be obtained with
        `results[docs_chunk_offsets[i][0]:docs_chunk_offsets[i][1]]`.
    :return Iterable[_TaskResult]: Results per document.
    """

extract(docs)

Extract all values from doc instances that are to be injected into the prompts.

Parameters:

Name Type Description Default
docs Iterable[Doc]

Docs to extract values from.

 required

Returns:

Type Description
Iterable[dict[str, Any]]

All values from doc instances that are to be injected into the prompts
Source code in sieves/tasks/predictive/bridges.py 
118
119
120
121
122
123
124
def extract(self, docs: Iterable[Doc]) -> Iterable[dict[str, Any]]:
    """Extract all values from doc instances that are to be injected into the prompts.

    :param docs: Docs to extract values from.
    :return Iterable[dict[str, Any]]: All values from doc instances that are to be injected into the prompts
    """
    return ({"text": doc.text if doc.text else None} for doc in docs)

integrate(results, docs) abstractmethod

Integrate results into Doc instances.

Parameters:

Name Type Description Default
results Iterable[TaskResult]

Results from prompt executable.

 required
docs Iterable[Doc]

Doc instances to update.

 required

Returns:

Type Description
Iterable[Doc]

Updated doc instances.
Source code in sieves/tasks/predictive/bridges.py 
126
127
128
129
130
131
132
133
@abc.abstractmethod
def integrate(self, results: Iterable[TaskResult], docs: Iterable[Doc]) -> Iterable[Doc]:
    """Integrate results into Doc instances.

    :param results: Results from prompt executable.
    :param docs: Doc instances to update.
    :return Iterable[Doc]: Updated doc instances.
    """