nlphose
flair
nlphose | flair | |
---|---|---|
4 | 9 | |
10 | 13,582 | |
- | 0.5% | |
2.7 | 9.4 | |
over 2 years ago | 2 days ago | |
Jupyter Notebook | Python | |
Apache License 2.0 | GNU General Public License v3.0 or later |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
nlphose
-
NlphoseBuilder : A tool to create NLP pipelines via drag and drop
The tool generates a nlphose command that can be executed in a docker container to run the pipeline. These pipelines can process streaming text like tweets or static data like files. They can be executed just like normal shell command using nlphose. Let me show you what I mean !
-
Create NLP pipelines with drag and drop
Recently I have started work on query builder GUI for my open source project nlphose.
- nlphose is a collection of command line utilities, which can be piped together to create complex NLP pipelines for processing stream of tweets (or any other textual data). Currently supports sentiment analysis, 0-shot classification, Q&A, NER, Chunking.
- nlphose : A collection of utilities, which can be piped together to create complex NLP pipelines for processing tweets (and other data); inspired by the “Unix tools philosophy”. Currently supports sentiment analysis, question answering , zero-shot classification, language detection, NER, chunking
flair
- Flair: A simple framework for state-of-the-art Natural Language Processing
-
Artificial Intelligence sentiment analysis of the Harry Potter movies. The greener the edge the happier the conversations, the bigger the edge the more they talk. Made by me.
The code of the module is available there for easy access: https://github.com/flairNLP/flair
-
The Spacy NER model for Spanish is terrible
Had the same experience with the german model in spacy (but tbh, the quailty of my textdata was bad). A bert based approach with flair really improved my results. I think there is a spanish pretrained model also available
-
How to create a dataset for training NER models when you only have entity data
We have a list of entities in text files separated with a new line. We intend to train the flair model to detect these entities in text, but NER models require the entity to be labeled in a paragraph with BOI format.
-
Preparing data for training NER models
Training most of the Named Entity Recognition (NER) models for example Flair usually needs to format data in BOI tagging) scheme as shown below where each sentence is separated by blank line
-
German POS Corpus for Commercial use
I had the same problem a couple years ago. I think Flair, form Zalando uses a different Corpus. However, it's not great and I am pretty sure they are infringing the license anyway...
-
Advice for how to approach classifying apartment posts on facebook?
For example, my first approach to the pet sentences would be to label all sentences within a respective text corpus containing according information for either yes or no. You would then convert this to a tertiary tag set, something like ["pet allowed", "pet not allowed", "irrelevant"]. You could then try out a model based on SentenceBert, other sentence-level embeddings/language models or 1D CNNs for this. flairNLP (https://github.com/flairNLP/flair) is a small, little framework which provides comfortable high-level access to different common language models which integrates perfectly with pyTorch.
-
SpaCy VS Transformers for NER
For NER, if you don't need the full toolkit of spacy, I'd highly recommend checking out Flair. It will likely run faster than transformer-based models (like en_core_web_trf) and it tends to be one of the best performing approaches to NER.
-
[D] NLP Q: How to extract this part from a messy short text?
You then train the whole thing on sequences where each position has a label that is begin/inside/outside and thus you can calculate cross-entropy loss. So all in all it is basically: https://github.com/flairNLP/flair, https://huggingface.co/transformers/model_doc/distilbert.html#tfdistilbertforsequenceclassification or any huggingface model "for sequence classificaiton" or but just char based instead of word based. The CRF layer (as included in flair) is optional but may be useful.
What are some alternatives?
ABSA-PyTorch - Aspect Based Sentiment Analysis, PyTorch Implementations. 基于方面的情感分析,使用PyTorch实现。
spacy-models - 💫 Models for the spaCy Natural Language Processing (NLP) library
nlphoseGUI - This tool allows you to create Natural Language Processing pipelines for use with nlphose using a Blockly based GUI editor in any browser. As you create a pipeline it shows you the corresponding nlphose command which will execute the pipeline.
BERT-NER - Pytorch-Named-Entity-Recognition-with-BERT
blockly - The web-based visual programming editor.
spaCy - 💫 Industrial-strength Natural Language Processing (NLP) in Python
wink-eng-lite-model - English lite language model for wink-nlp.
Stanza - Stanford NLP Python library for tokenization, sentence segmentation, NER, and parsing of many human languages
awesome-sentiment-analysis - Repository with all what is necessary for sentiment analysis and related areas
gensim - Topic Modelling for Humans
FinBERT - A Pretrained BERT Model for Financial Communications. https://arxiv.org/abs/2006.08097
seqeval - A Python framework for sequence labeling evaluation(named-entity recognition, pos tagging, etc...)