Semantic graph uses top most valuable phrases and terms of given text to build a mindmap-like associative graph of entities.
Our services are built on top of modern async programming models for serving HTTP requests and communicating with processing queues. It allows services to be high load ready and able to be scaled horizontally in a very efficient way even on the fly. Technology stack: Python 3.4, asyncio, aiohttp, aiozmq (ZeroMQ, 0mq), GoLang, Scala, and C/C++ in rare cases for computational bound bottlenecks.
We develop Natural Language Processing systems using existing tools as well as our own GoLang based implementation called nlp4go text classifiers, POS taggers, Sentence Parsing, Spell Checkers, Integrating NLP into Recommender System cores and use it for feature extraction to improve performance of ML algorithms. Recommender Systems building experience.
Machine learning approaches we use: Search for a best model using Cross-Validation and Grid Search; Lots of ML models like SVM, Bayesian Classifiers, GBM, GBR, Logistic Regression, Random Forests, etc. Technology stack: Python scikit-learn, Spark MLLib.
Text data classification and clusterization, information retrieval, digging unstructured data. We use regular approaches such as TF-IDF, bag of words, Bayesian and SVM classifiers as well as results of our research regarding to feature extraction and NLP tricks for dimensionality reduction and clusterization. Tools set: NLTK, Go-Nlp-Tools, Scalanlp.
All those popular words like BIG DATA, HIGH LOAD, FUNCTIONAL PROGRAMMING and CLOUD COMPUTATIONS are very close to our work, but we won’t speak a lot of them, we just use all this stuff.