🐠^🐠 NEMO² - Neural Modeling for (Hebrew) Named Entities and Morphology

Table of Contents

• Introduction
• Main Features
• Setup
• Basic Usage
• Models and Scenarios
• Important Notes
• Training Your Own Model
• Morpheme and Word Embeddings
• Evaluation
• Ben-Mordecai Corpus
• Citations

Introduction

Code and models for neural modeling of Hebrew NER. Described in the TACL paper "Neural Modeling for Named Entities and Morphology (NEMO²)" along with extensive experiments on the different modeling scenarios provided in this repository.

Main Features

1. Trained on the Hebrew NER and Morphology NEMO corpus of gold annotated Modern Hebrew news articles.
2. Multiple modeling options to go from raw Hebrew text to morpheme and/or token-level NER boundaries.
3. Neural model implementation of NCRF++
4. bclm is used for reading and transforming morpho-syntactic information layers.

Setup

Pre-requisites:

1. Clone this NEMO repo: git clone https://github.com/OnlpLab/NEMO.git
2. Enter the repo directory: cd NEMO
3. Preferably in a virtual env: pip install -r requirements.txt
4. Unpack model files: gunzip data/*.gz
5. Install yap: https://github.com/OnlpLab/yap

To run API server

1. In YAP folder, run YAP API server ./yap api (if you specify a port, change it in config.py)
2. In NEMO folder, run NEMO API server uvicorn api_main:app --reload --port 8090

To run on file input (CLI): nemo.py

1. Change YAP_PATH in config.py to the path of your local yap executable.

Basic Usage

API Usage

1. Once the API server is up, check out the API documentation by opening (http://localhost:8090/docs) in your browser.
2. You can find the available API endpoints and more usage examples in api_usage.ipynb.

File Input Usage (CLI)

1. All you need to do is run nemo.py with a specific command (scenario), on a text file of Hebrew sentences separated by a line-break.
2. You can run a neural NER model directly, or choose a full end-to-end scenario that includes morphological segmentation and alignments (described fully in the next section). e.g.:
   • the run_ner_model command with the token-single model will tokenize sentences and run the token-single model:
     • python nemo.py run_ner_model token-single example.txt example_output.txt
   • the morph_hybrid command runs the end-to-end segmentation and NER pipeline which provided our best performing morpheme-level NER boundaries:
     • python nemo.py morph_yap morph example.txt example_output_MORPH.txt
3. You can find outputs of different commands on the input in example.txt in: morph_hybrid_align_tokens, morph_hybrid, morph_yap, multi_align_hybrid, single
4. For a full list of the available commands please consult the next section and the inline documentation at the end of nemo.py.

Models and Scenarios

Models are all standard Bi-LSTM-CRF with char encoding (LSTM/CNN) of NCRFpp with pre-trained fastText embeddings. Differences between models lay in:

1. Input units: morphemes morph vs. tokens token-*
2. Output label set: token-single single sequence labels (e.g. B-ORG) vs. token-multi multi-labels (atomic labels, e.g. O-ORG^B-ORG^I-ORG) that predict, in order, the labels for the morphemes the token is made of.

Token-based Models	Morpheme-based Model

Morphemes must be predicted. This is done by performing morphological disambiguation (MD). We offer two options to do so:

1. Standard pipeline: MD using YAP. This is used in the morph_yap command, which runs our morph NER model on the output of YAP joint segmentation.
2. Hybrid pipeline: MD using our best performing Hybrid approach, which uses the output of the token-multi model to reduce the MD option space. This is used in morph_hybrid, multi_align_hybrid and morph_hybrid_align_tokens. We will explain these scenarios next.

MD Approach	Commands
Standard	morph_yap
Hybrid	morph_hybrid, multi_align_hybrid, morph_hybrid_align_tokens

Finally, to get our desired output (tokens/morphemes), we can choose between different scenarios, some involving extra post-processing alignments:

1. To get morpheme-level labels we have two options:
   • Run our morph NER model on predicted morphemes: Commands: morph_yap or morph_hybrid (better).
   • token-multi labels can be aligned with predicted morphemes to get morpheme-level boundaries. Command: multi_align_hybrid.

Run morph NER on Predicted Morphemes	Multi Predictions Aligned with Predicted Morpheme
morph_yap,morph_hybrid	multi_align_hybrid

2. To get token-level labels we have three options:
   • run_ner_model command with token-single model.
   • the predicted labels of the token-multi can be mapped to token-single labels to get standard token-single output. The command multi_to_single does this end-to-end.
   • Morpheme-level output can be aligned back to token-level boundaries. Command: morph_hybrid_align_tokens (this achieved best token-level results in our experiments).

Run token-single	Map token-multi to token-single	Align morph NER with Tokens
run_ner_model token-single	multi_to_single	morph_hybrid_align_tokens

• Note: while the morph_hybrid* scenarios offer the best performance, they are less efficient since they requires running both morph and token-multi NER models.

Important Notes

1. NCRFpp was great for our experiments on the NEMO corpus (which is given, constant, data), but it holds some caveats for real life scenarios of arbitrary text:
   • fastText is not used on the fly to obtain vectors for OOV words (i.e. those that were not seen in our Wikipedia corpus). Instead, it is used as a regular embedding matrix. Hence the full generalization capacities of fastText, as shown in our experiments, are not available in the currently provided models, which will perform slightly worse than they could on arbitrary text. In our experiments we created such a matrix in advance with all the words in the NEMO corpus and used it during training. Information regarding training your own model with your own vocabulary in the next section.
   • If you do wish to replicate our reported results on the Hebrew treebank, download the *oov* models from here and extract to the data/ folder (they already appear in config.py).
2. In the near future we plan to publish a cleaner end-to-end implementation, including use of our new AlephBERT pre-trained Transformer models.
3. For archiving and reproducibility purposes, our original code used for experiments and analysis can be found in the following repos: https://github.com/cjer/NCRFpp, https://github.com/cjer/NER (beware - 2 years of Jupyter notebooks).

Training your own model

We provide template NCRF++ config files. These files already contain the hyperparameters we used in our training. To train your own model:

1. Copy the config for the variant (token-multi, token-single, morph) you wish to use from the ncrf_train_configs folder.
2. Change the parameter word_emb_dir to that of an embedding vectors file in standard word2vec textual format. You can use the fastText bin models we make available (in the next section) or any other embedding vectors of your choice.
3. Run the following in your shell:

python ncrf_main.py --config <path_to_config> --device <gpu_device_number>

4. For more information, please consult NCRF++ documentation.
5. To evaluate your trained models, please consult the evaluation section.

Morpheme and Word Embeddings

The word embeddings we trained and used in our models are available:

1. Space-delimited tokens (traditional word embeddings): fastText (bin, text), GloVe, word2vec
2. Morphemes: fastText (bin, text), GloVe, word2vec

These were trained on a 2013 Wiki dump corpus by Yoav Goldberg, which we re-tokenized and then re-parsed using YAP:

1. Space-delimited tokens
2. Morphemes, automatic YAP segmentation (using the morpheme FORM as the unit for embedding)
3. CONLL files of full morpho-syntactic output of YAP

Evaluation

To evaluate your predictions against gold use the ne_evaluate_mentions.py script. Evaluation looks for exact match of string and entity category, but is slightly different than the standard CoNLL2003 evaluation commonly used for NER. The reason is that predicted segmentation differs from gold, so positional indexes of sequence labels cannot be used. What we do instead, is extract multi-sets of entity mentions and use set operations to compute precision, recall and F1-score. You can find more detailed discussion of evaluation in the NEMO² paper.

To evaluate an output prediction file against a gold file use:

python ne_evaluate_mentions.py <path_to_gold_ner> <path_to_predicted_ner>

If you're within python, just call ne_evaluate_mentions.evaluate_files(...) with the same parameters.

Ben-Mordecai Corpus

In our NEMO² paper we also evaluate our models on the Ben-Mordecai Hebrew NER Corpus (BMC). The 3 random splits we used can be found here.

Citations

If you use any of the NEMO² code, models, embeddings or the NEMO corpus, please cite the NEMO² paper:

@article{DBLP:journals/corr/abs-2007-15620,
  author    = {Dan Bareket and
               Reut Tsarfaty},
  title     = {Neural Modeling for Named Entities and Morphology (NEMO{\^{}}2)},
  journal   = {CoRR},
  volume    = {abs/2007.15620},
  year      = {2020},
  url       = {https://arxiv.org/abs/2007.15620},
  archivePrefix = {arXiv},
  eprint    = {2007.15620},
  timestamp = {Mon, 03 Aug 2020 14:32:13 +0200},
  biburl    = {https://dblp.org/rec/journals/corr/abs-2007-15620.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}

If you use the NEMO²'s NER models please also cite NCRF++:

@inproceedings{yang2018ncrf,  
 title={{NCRF}++: An Open-source Neural Sequence Labeling Toolkit},  
 author={Yang, Jie and Zhang, Yue},  
 booktitle={Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics},
 Url = {http://aclweb.org/anthology/P18-4013},
 year={2018}  
}