Skip to content

2023-10-06-finembedding_e5_base_en #685

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
Oct 6, 2023
Merged

2023-10-06-finembedding_e5_base_en #685

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
135c844
Add model 2023-10-06-finembedding_e5_base_en
dcecchini Oct 6, 2023
73565e0
Add model 2023-10-06-finner_absa_sm_en
dcecchini Oct 6, 2023
1125c39
Add model 2023-10-06-finassertion_absa_sm_en
dcecchini Oct 6, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
151 changes: 151 additions & 0 deletions docs/_posts/dcecchini/2023-10-06-finassertion_absa_sm_en.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,151 @@
---
layout: model
title: Financial Assertion of Sentiment (sm, Small)
author: John Snow Labs
name: finassertion_absa_sm
date: 2023-10-06
tags: [finance, assertion, en, sentiment_analysis, licensed]
task: Assertion Status
language: en
edition: Finance NLP 1.0.0
spark_version: 3.0
supported: true
annotator: AssertionDLModel
article_header:
type: cover
use_language_switcher: "Python-Scala-Java"
---

## Description

This assertion model classifies financial entities into a sentiment. It is designed to be used together with the associated NER model.

## Predicted Entities

`POSITIVE`, `NEGATIVE`, `NEUTRAL`

{:.btn-box}
<button class="button button-orange" disabled>Live Demo</button>
<button class="button button-orange" disabled>Open in Colab</button>
[Download](https://s3.amazonaws.com/auxdata.johnsnowlabs.com/finance/models/finassertion_absa_sm_en_1.0.0_3.0_1696606845902.zip){:.button.button-orange.button-orange-trans.arr.button-icon.hidden}
[Copy S3 URI](s3://auxdata.johnsnowlabs.com/finance/models/finassertion_absa_sm_en_1.0.0_3.0_1696606845902.zip){:.button.button-orange.button-orange-trans.button-icon.button-copy-s3}

## How to use



<div class="tabs-box" markdown="1">
{% include programmingLanguageSelectScalaPythonNLU.html %}
```python
documentAssembler = (
nlp.DocumentAssembler().setInputCol("text").setOutputCol("document")
)

# Sentence Detector annotator, processes various sentences per line
sentenceDetector = (
nlp.SentenceDetector()
.setInputCols(["document"])
.setOutputCol("sentence")
)

# Tokenizer splits words in a relevant format for NLP
tokenizer = (
nlp.Tokenizer().setInputCols(["sentence"]).setOutputCol("token")
)

bert_embeddings = (
nlp.BertEmbeddings.pretrained("bert_embeddings_sec_bert_base", "en")
.setInputCols("document", "token")
.setOutputCol("embeddings")
.setMaxSentenceLength(512)
)

clinical_ner = (
finance.NerModel.pretrained("finner_absa_sm", "en", "finance/models")
.setInputCols(["sentence", "token", "embeddings"])
.setOutputCol("ner")
)

ner_converter = (
finance.NerConverterInternal()
.setInputCols(["sentence", "token", "ner"])
.setOutputCol("ner_chunk")
)

assertion_model = (
finance.AssertionDLModel.pretrained("finassertion_absa_sm", "en", "finance/models")
.setInputCols(["sentence", "ner_chunk", "embeddings"])
.setOutputCol("assertion")
)

nlpPipeline = nlp.Pipeline(
stages=[
documentAssembler,
sentenceDetector,
tokenizer,
bert_embeddings,
clinical_ner,
ner_converter,
assertion_model,
]
)


text = "Equity and earnings of affiliates in Latin America increased to $4.8 million in the quarter from $2.2 million in the prior year as the commodity markets in Latin America remain strong through the end of the quarter."

spark_df = spark.createDataFrame([[text]]).toDF("text")

result = model.fit(spark_df ).transform(spark_df)

result.select(
F.explode(
F.arrays_zip("ner_chunk.result", "ner_chunk.metadata")
).alias("cols")
).select(
F.expr("cols['0']").alias("entity"),
F.expr("cols['1']['entity']").alias("label"),
).show(
50, truncate=False
)
```

</div>

## Results

```bash
+--------+---------+
|entity |label |
+--------+---------+
|Equity |LIABILITY|
|earnings|PROFIT |
+--------+---------+
```

{:.model-param}
## Model Information

{:.table-model}
|---|---|
|Model Name:|finassertion_absa_sm|
|Compatibility:|Finance NLP 1.0.0+|
|License:|Licensed|
|Edition:|Official|
|Input Labels:|[document, chunk, embeddings]|
|Output Labels:|[assertion]|
|Language:|en|
|Size:|2.7 MB|

## References

In-house annotations of earning call transcripts.

## Benchmarking

```bash
label precision recall f1-score support

NEGATIVE 0.57 0.42 0.48 74
NEUTRAL 0.51 0.70 0.59 184
POSITIVE 0.75 0.64 0.69 324
```
93 changes: 93 additions & 0 deletions docs/_posts/dcecchini/2023-10-06-finembedding_e5_base_en.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,93 @@
---
layout: model
title: Finance E5 Embedding Base
author: John Snow Labs
name: finembedding_e5_base
date: 2023-10-06
tags: [finance, en, licensed, e5, sentence_embedding, onnx]
task: Embeddings
language: en
edition: Finance NLP 1.0.0
spark_version: 3.0
supported: true
engine: onnx
annotator: E5Embeddings
article_header:
type: cover
use_language_switcher: "Python-Scala-Java"
---

## Description

This model is a financial version of the E5 base model fine-tuned on earning call transcripts and finance question-answering datasets. Reference: Wang, Liang, et al. "Text embeddings by weakly-supervised contrastive pre-training." arXiv preprint arXiv:2212.03533 (2022).

## Predicted Entities



{:.btn-box}
<button class="button button-orange" disabled>Live Demo</button>
<button class="button button-orange" disabled>Open in Colab</button>
[Download](https://s3.amazonaws.com/auxdata.johnsnowlabs.com/finance/models/finembedding_e5_base_en_1.0.0_3.0_1696603847700.zip){:.button.button-orange.button-orange-trans.arr.button-icon.hidden}
[Copy S3 URI](s3://auxdata.johnsnowlabs.com/finance/models/finembedding_e5_base_en_1.0.0_3.0_1696603847700.zip){:.button.button-orange.button-orange-trans.button-icon.button-copy-s3}

## How to use



<div class="tabs-box" markdown="1">
{% include programmingLanguageSelectScalaPythonNLU.html %}
```python
document_assembler = (
nlp.DocumentAssembler().setInputCol("text").setOutputCol("document")
)

E5_embedding = (
nlp.E5Embeddings.pretrained(
"finembedding_e5_base", "en", "finance/models"
)
.setInputCols(["document"])
.setOutputCol("E5")
)
pipeline = nlp.Pipeline(stages=[document_assembler, E5_embedding])

data = spark.createDataFrame(
[["What is the best way to invest in the stock market?"]]
).toDF("text")

result = pipeline.fit(data).transform(data)
result. Select("E5.result").show()
```

</div>

## Results

```bash
+----------------------------------------------------------------------------------------------------+
| embeddings|
+----------------------------------------------------------------------------------------------------+
|[0.45521045, -0.16874692, -0.06179046, -0.37956607, 1.152633, 0.6849592, -0.9676384, 0.4624033, ...|
+----------------------------------------------------------------------------------------------------+
```

{:.model-param}
## Model Information

{:.table-model}
|---|---|
|Model Name:|finembedding_e5_base|
|Compatibility:|Finance NLP 1.0.0+|
|License:|Licensed|
|Edition:|Official|
|Input Labels:|[document]|
|Output Labels:|[E5]|
|Language:|en|
|Size:|398.5 MB|

## References

For our Finance models, we will use publicly available datasets to fine-tune the model:

- [FiQA](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/)
- In-house annotated Earning Calls Transcripts
147 changes: 147 additions & 0 deletions docs/_posts/dcecchini/2023-10-06-finner_absa_sm_en.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,147 @@
---
layout: model
title: Financial NER for Aspect-based Sentiment Analysis (sm, Small)
author: John Snow Labs
name: finner_absa_sm
date: 2023-10-06
tags: [finance, en, ner, licensed]
task: Named Entity Recognition
language: en
edition: Finance NLP 1.0.0
spark_version: 3.0
supported: true
annotator: FinanceNerModel
article_header:
type: cover
use_language_switcher: "Python-Scala-Java"
---

## Description

This NER model identifies entities that can be associated with a financial sentiment. The model is designed to be used with the associated Assertion Status model that classifies the entities into a sentiment category.

## Predicted Entities

`REVENUE`, `EXPENSE`, `PROFIT`, `KPI`, `GAINS`, `ASSET`, `LIABILITY`, `CASHFLOW`, `LOSSES`, `FREE_CASH_FLOW`

{:.btn-box}
<button class="button button-orange" disabled>Live Demo</button>
<button class="button button-orange" disabled>Open in Colab</button>
[Download](https://s3.amazonaws.com/auxdata.johnsnowlabs.com/finance/models/finner_absa_sm_en_1.0.0_3.0_1696605316183.zip){:.button.button-orange.button-orange-trans.arr.button-icon.hidden}
[Copy S3 URI](s3://auxdata.johnsnowlabs.com/finance/models/finner_absa_sm_en_1.0.0_3.0_1696605316183.zip){:.button.button-orange.button-orange-trans.button-icon.button-copy-s3}

## How to use



<div class="tabs-box" markdown="1">
{% include programmingLanguageSelectScalaPythonNLU.html %}
```python
document_assembler = nlp.DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")

sentence_detector = nlp.SentenceDetector() \
.setInputCols(["document"]) \
.setOutputCol("sentence") \
.setCustomBounds(["\n\n"])

tokenizer = nlp.Tokenizer()\
.setInputCols(["sentence"])\
.setOutputCol("token")

embeddings = nlp.BertEmbeddings.pretrained("bert_embeddings_sec_bert_base","en")\
.setInputCols(["sentence", "token"])\
.setOutputCol("embeddings")\
.setCaseSensitive(True)\
.setMaxSentenceLength(512)

ner_model = finance.NerModel.pretrained("finner_absa_sm", "en", "finance/models")\
.setInputCols(["sentence", "token", "embeddings"])\
.setOutputCol("ner")\

ner_converter = finance.NerConverterInternal()\
.setInputCols(["sentence", "token", "ner"])\
.setOutputCol("ner_chunk")

pipeline = nlp.Pipeline(stages=[
document_assembler,
sentence_detector,
tokenizer,
embeddings,
ner_model,
ner_converter
])

model = pipeline.fit(spark.createDataFrame([[""]]).toDF("text"))


text = "Equity and earnings of affiliates in Latin America increased to $4.8 million in the quarter from $2.2 million in the prior year as the commodity markets in Latin America remain strong through the end of the quarter."

spark_df = spark.createDataFrame([[text]]).toDF("text")

result = model. Transform(spark_df)
result. Select(F.explode(F.arrays_zip('ner_chunk.result', 'ner_chunk.metadata')).alias("cols")) \
.select(F.expr("cols['0']").alias("entity"),
F.expr("cols['1']['entity']").alias("label")).show(50, truncate = False)

```

</div>

## Results

```bash
+--------+---------+
|entity |label |
+--------+---------+
|Equity |LIABILITY|
|earnings|PROFIT |
+--------+---------+
```

{:.model-param}
## Model Information

{:.table-model}
|---|---|
|Model Name:|finner_absa_sm|
|Compatibility:|Finance NLP 1.0.0+|
|License:|Licensed|
|Edition:|Official|
|Input Labels:|[sentence, token, embeddings]|
|Output Labels:|[ner]|
|Language:|en|
|Size:|16.3 MB|

## References

In-house annotations of earning call transcripts.

## Benchmarking

```bash
label precision recall f1-score support

B-ASSET 0.6000 0.2400 0.3429 25
B-CASHFLOW 0.7000 0.5833 0.6364 12
B-EXPENSE 0.7222 0.6500 0.6842 60
B-FREE_CASH_FLOW 1.0000 1.0000 1.0000 8
B-GAINS 0.7333 0.5946 0.6567 37
B-KPI 0.7143 0.5556 0.6250 36
B-LIABILITY 0.5000 0.2778 0.3571 18
B-LOSSES 0.7143 0.7143 0.7143 7
B-PROFIT 0.8462 0.8919 0.8684 37
B-REVENUE 0.7385 0.8000 0.7680 60
I-ASSET 0.8000 0.3636 0.5000 11
I-CASHFLOW 0.9091 0.9091 0.9091 11
I-EXPENSE 0.7451 0.6230 0.6786 61
I-FREE_CASH_FLOW 1.0000 1.0000 1.0000 17
I-GAINS 0.8333 0.6667 0.7407 30
I-KPI 0.8500 0.5000 0.6296 34
I-LIABILITY 0.5000 0.5000 0.5000 6
I-LOSSES 0.7143 0.6250 0.6667 8
I-PROFIT 0.8621 0.9615 0.9091 26
I-REVENUE 0.7600 0.7308 0.7451 26
O 0.9839 0.9923 0.9880 8660
```