Description
This model is a Named Entity Recognition (NER) model focused on financial numeric items. It identifies 7 numeric financial entities from diverse 10-Q and 10-K reports. These entities are annotated using eXtensible Business Reporting Language (XBRL) tags. The annotation process primarily targets numerical tokens, and the context plays a crucial role in accurately assigning the appropriate entity type from the 139 most common financial entities available in the dataset.
Predicted Entities
NumberOfReportableSegments, RevenueFromContractWithCustomerExcludingAssessedTax, ContractWithCustomerLiabilityRevenueRecognized, CumulativeEffectOfNewAccountingPrincipleInPeriodOfAdoption, NumberOfOperatingSegments, RevenueFromRelatedParties, RevenueFromContractWithCustomerIncludingAssessedTax
How to use
documentAssembler = nlp.DocumentAssembler() \
.setInputCol("text") \
.setOutputCol("document")
sentence = nlp.SentenceDetector() \
.setInputCols(["document"]) \
.setOutputCol("sentence")
tokenizer = nlp.Tokenizer()\
.setInputCols(["sentence"])\
.setOutputCol("token")\
.setContextChars(['.', ',', ';', ':', '!', '?', '*', '-', '(', ')', '”', '’', '$','€'])
embeddings = nlp.BertEmbeddings.pretrained("bert_embeddings_sec_bert_base","en") \
.setInputCols(["document", "token"]) \
.setOutputCol("embeddings")\
.setMaxSentenceLength(512)
nerTagger = finance.NerModel.pretrained('finner_10q_xbrl_lg_revenue', 'en', 'finance/models')\
.setInputCols(["sentence", "token", "embeddings"])\
.setOutputCol("ner")
pipeline = nlp.Pipeline(stages=[documentAssembler,
sentence,
tokenizer,
embeddings,
nerTagger
])
text = "In addition , during the first quarter of 2016 , the Company made the decision to divide the Southeast Florida operating division into two operating segments to maximize operational efficiencies given the continued growth of the division ."
df = spark.createDataFrame([[text]]).toDF("text")
fit = pipeline.fit(df)
result = fit.transform(df)
result_df = result.select(F.explode(F.arrays_zip(result.token.result,result.ner.result, result.ner.metadata)).alias("cols"))\
.select(F.expr("cols['0']").alias("token"),\
F.expr("cols['1']").alias("ner_label"),\
F.expr("cols['2']['confidence']").alias("confidence"))
result_df.show(50, truncate=100)
Results
+------------+---------------------------+----------+
|token |ner_label |confidence|
+------------+---------------------------+----------+
|In |O |1.0 |
|addition |O |1.0 |
|, |O |1.0 |
|during |O |1.0 |
|the |O |1.0 |
|first |O |1.0 |
|quarter |O |1.0 |
|of |O |1.0 |
|2016 |O |1.0 |
|, |O |1.0 |
|the |O |1.0 |
|Company |O |1.0 |
|made |O |1.0 |
|the |O |1.0 |
|decision |O |1.0 |
|to |O |1.0 |
|divide |O |1.0 |
|the |O |1.0 |
|Southeast |O |1.0 |
|Florida |O |1.0 |
|operating |O |1.0 |
|division |O |1.0 |
|into |O |1.0 |
|two |B-NumberOfOperatingSegments|0.9831 |
|operating |O |1.0 |
|segments |O |1.0 |
|to |O |1.0 |
|maximize |O |1.0 |
|operational |O |1.0 |
|efficiencies|O |1.0 |
|given |O |1.0 |
|the |O |1.0 |
|continued |O |1.0 |
|growth |O |1.0 |
|of |O |1.0 |
|the |O |1.0 |
|division |O |1.0 |
|. |O |1.0 |
+------------+---------------------------+----------+
Model Information
| Model Name: | finner_10q_xbrl_lg_revenue |
| 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
An in-house modified version of https://huggingface.co/datasets/nlpaueb/finer-139, re-splited and filtered to focus on sentences with bigger density of tags.
Benchmarking
label precision recall f1-score support
B-ContractWithCustomerLiabilityRevenueRecognized 0.9516 0.9502 0.9509 642
B-CumulativeEffectOfNewAccountingPrincipleInPeriodOfAdoption 0.8588 0.9648 0.9087 227
B-NumberOfOperatingSegments 0.8159 0.8654 0.8399 379
B-NumberOfReportableSegments 0.9186 0.8899 0.9040 672
B-RevenueFromContractWithCustomerExcludingAssessedTax 0.7234 0.9223 0.8108 618
B-RevenueFromContractWithCustomerIncludingAssessedTax 0.9130 0.2500 0.3925 168
B-RevenueFromRelatedParties 0.9161 0.9192 0.9176 594
O 0.9998 0.9988 0.9993 80111
accuracy - - 0.9942 83411
macro-avg 0.8872 0.8451 0.8405 83411
weighted-avg 0.9947 0.9942 0.9940 83411