TeXoo is a framework for Deep Learning based text analytics in Java developed at DATEXIS, Beuth University of Applied Sciences Berlin. TeXoo comes with a NLP-style document model and a zoo of Deep Learning extraction models which you can access in texoo-models module. Here is a brief overview:
- Java Framework for language-independent text extraction
- Language-independent document model
- Convenient document readers with tokenization and sentence splitting
- Named Entity Recognition
- Named Entity Linking
- Topic Classification and Segmentation
These instructions will get you a copy of TeXoo up and running on your local machine for development and testing purposes. If you are going to use TeXoo as a Maven dependency only, you might skip this step.
TeXoo comes with a Dockerfile that contains all software necessary to run on most systems, including the CUDA 10 toolkit for GPUs.
-
Docker Container platform
https://docs.docker.com/install/ -
nvidia-docker v2 for CUDA 10 support
https://github.com/nvidia/nvidia-docker/wiki/Installation-(version-2.0)
The following dependencies are required if you are planning to run TeXoo locally. They are already contained in the Dockerfile:
- OpenJDK 8
- Apache Maven Build system for Java
https://maven.apache.org/guides/index.html
First we need to build a docker image with all dependencies (including CUDA 10.1):
- run
docker build -t texoo .
And then we're ready to build TeXoo from source:
- run
bin/run-docker texoo-build
There exist several run scripts in the bin/ directory. You can start them right in the docker container, e.g. run all JUnit tests:
- run
bin/run-docker texoo-testor - runbin/run-docker-cuda texoo-test
See the Modules Overview for more examples.
To use TeXoo NER in your Java project, just add the following dependencies to your pom.xml:
<dependency>
<groupId>de.datexis</groupId>
<artifactId>texoo-core</artifactId>
<version>1.3.3</version>
<type>jar</type>
</dependency>
<dependency>
<groupId>de.datexis</groupId>
<artifactId>texoo-entity-recognition</artifactId>
<version>1.3.3</version>
<type>jar</type>
</dependency>
To enable CUDA support, add the following dependencies in your project:
<dependency>
<groupId>org.nd4j</groupId>
<artifactId>nd4j-cuda-9.2-platform</artifactId>
<version>${dl4j.version}</version>
</dependency>
<!-- DL4j cuDNN -->
<dependency>
<groupId>org.deeplearning4j</groupId>
<artifactId>deeplearning4j-cuda-9.2</artifactId>
<version>${dl4j.version}</version>
</dependency>
<!-- DL4j CUDA + cuDNN binaries -->
<dependency>
<groupId>org.bytedeco.javacpp-presets</groupId>
<artifactId>cuda</artifactId>
<version>9.2-7.1-1.4.2</version>
<classifier>linux-x86_64-redist</classifier>
</dependency>
And to enable AVX512 CPU optimizations, add the following dependencies in your project:
<dependency>
<groupId>org.nd4j</groupId>
<artifactId>nd4j-native</artifactId>
<classifier>linux-x86_64-avx512</classifier>
</dependency>
See the examples module for some implementation examples.
texoo-core – Document Model and Core Library
| Package / Class | Description |
|---|---|
| de.datexis.model | TeXoo Document model (see below) |
| de.datexis.encoder | Implementations of Bag-of-words, Word2Vec, Trigrams, etc. |
| DocumentFactory | Factory to create Document objects from text |
| RawTextDatasetReader | Reader to create Datasets from files |
| AnnotatorFactory | Factory to create and load models from the zoo |
| ObjectSerializer | Helper methods to import/export JSON |
texoo-entity-recognition – Named Entity Recognition (NER)
This module contains Annotators for Named Entity Recognition (NER). This is a very robust deep learning model that can be trained with only 4000-5000 sentences. It is based on a bidirection LSTM with Letter-trigram encoding, see http://arxiv.org/abs/1608.06757.
- run
bin/run-docker texoo-annotate-ner
usage: texoo-annotate-ner -i <arg> [-o <arg>]
TeXoo: run pre-trained MentionAnnotator model
-i,--input <arg> path or file name for raw input text
-o,--output <arg> path to create and store the output JSON, otherwise dump to stdout
- run
bin/run-docker texoo-train-ner
usage: texoo-train-ner -i <arg> [-l <arg>] -o <arg> [-t <arg>] [-u] [-v
<arg>]
TeXoo: train MentionAnnotator with CoNLL annotations
-i,--input <arg> path to input training data (CoNLL format)
-l,--language <arg> language to use for sentence splitting and
stopwords (EN or DE)
-o,--output <arg> path to create and store the model
-t,--test <arg> path to test data (CoNLL format)
-u,--ui enable training UI (http://127.0.0.1:9000)
-v,--validation <arg> path to validation data (CoNLL format)
- run
bin/run-docker texoo-train-ner-seed
usage: texoo-train-ner-seed -i <arg> -o <arg> -s <arg> [-u]
TeXoo: train MentionAnnotator with seed list
-i,--input <arg> path and file name pattern for raw input text
-o,--output <arg> path to create and store the model
-s,--seed <arg> path to seed list text file
-u,--ui enable training UI (http://127.0.0.1:9000)
| Package / Class | Description / Reference |
|---|---|
| MentionAnnotator | Named Entity Recognition |
| GenericMentionAnnotator | Pre-trained models for English and German |
| MatchingAnnotator | Gazetteer that uses Lists to annotate Documents |
| CoNLLDatasetReader | Reader for CoNLL files |
If you use this module for research, please cite:
Sebastian Arnold, Felix A. Gers, Torsten Kilias, Alexander Löser: Robust Named Entity Recognition in Idiosyncratic Domains. arXiv:1608.06757 [cs.CL] 2016 https://arxiv.org/abs/1608.06757
texoo-entity-linking – Named Entity Linking (NEL)
This module contains the Annotators for Named Entity Linking (NEL) (currently under development). There is no model included, but you can use the Knowledge Base and Annotators with your own datasets, see https://www.aclweb.org/anthology/C/C16/C16-2024.pdf.
| Package / Class | Description / Reference |
|---|---|
| NamedEntityAnnotator | Named Entity Linking used in TASTY |
| ArticleIndexFactory | Knowledge Base implemented as local Lucene Index which imports Wikidata entities |
If you use this module for research, please cite:
Sebastian Arnold, Robert Dziuba, Alexander Löser: TASTY: Interactive Entity Linking As-You-Type. COLING (Demos) 2016: 111–115
texoo-sector – Topic Classification and Segmentation (SECTOR)
Annotators for SECTOR models from WikiSection dataset.
| Package / Class | Description / Reference |
|---|---|
| SectorAnnotator | Topic Segmentation and Classification for Long Documents |
If you use this module for research, please cite:
Sebastian Arnold, Rudolf Schneider, Philippe Cudré-Mauroux, Felix A. Gers and Alexander Löser. SECTOR: A Neural Model for Coherent Topic Segmentation and Classification. Transactions of the Association for Computational Linguistics 2019 Vol. 7, 169-184
- run
bin/run-docker texoo-train-sector
usage: texoo-train-sector -i <arg> -o <arg> [-u]
TeXoo: train SectorAnnotator from WikiSection dataset
-i,--input <arg> file name of WikiSection training dataset
-o,--output <arg> path to create and store the model
-u,--ui enable training UI (http://127.0.0.1:9000)
- Deeplearning4j Machine learning library
http://deeplearning4j.org/documentation - ND4J Scientific computing library
http://nd4j.org/userguide - Apache OpenNLP Natural language processing
https://opennlp.apache.org/docs/
Sebastian Arnold – core developer https://prof.beuth-hochschule.de/loeser/people/sebastian-arnold/
Rudolf Schneider https://prof.beuth-hochschule.de/loeser/people/rudolf-schneider/
Copyright 2015-2020 Sebastian Arnold, Alexander Löser, Rudolf Schneider
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.