Will Google NL kill the market? Linguistic APIs review
Google has recently launched a new beta product – Google Cloud Natural Language API. As a rule, any new product by Google makes the company’s rivals get in a slight panic. What major players are to expect in the Natural Language area? Does it mean the end of free competition in the linguistic API market and dictatorship by “The good corporation”?
However, if we are to compose a list of other players in this sphere, a fear for Google’s dominance will disappear. Apart from major players, this top also features small companies that are notable for their flexibility and attention to customers’ needs. Let’s take a look at major API providers and mull over their peculiarities.
1) Alchemy IBM. AlchemyAPI's text analysis service started its existence in 2009. AlchemyLanguage provides for several API servers that analyze texts and define meta-data, including entities, sentiments, keywords, concepts, and relations between sentence components (nouns, verbs, subjects, objects, etc.), authors, and more. Beside English, Alchemy IBM supports other 7 languages (German, Spanish, French, Italian, Portuguese, Swedish, and Russian).
This Academy seems kind of an incubator for young and promising projects. Many companies started functioning after getting free access to this API within the Watson development platform. Nevertheless, a high price for its usage in commercial projects push many customers to search for alternative APIs.
2) Microsoft Cognitive Services based on Project Oxford and Bing and launched in March, 2016, offer a range of APIs for text analysis, including sentiment analysis (English, Spanish, French, and Portuguese), key phrases extraction (English, German, Spanish, and Japanese), detection of topics for a list of submitted text records (minimum 100) and language recognition.
3) Google Natural Language API. A beta version of this API service became available in July, 2016. For now, users can only relish basic tools: sentiment analysis (English), entity recognition, and syntax analysis (English, Spanish, and Japanese). The service is already available on the paid basis.
4) Intellexer API is a recently launched cloud semantic server (January, 2016). EffectiveSoft - software company (based in San Diego, USA, and Minsk, Belarus) has been dealing with elaborating semantic tools (Intellexer SDK) and custom solutions since 2003. For today, Intellexer offers a full range of tools to analyze English texts. The features Intellexer provides for can be checked via user-friendly demos. Except the API, another semantic product, Intellexer Summarizer, is available in the user’s account. Intellexer also is available in form of SDK and some desktop and mobile Apps.
5) TextRazor (created in London, UK, in 2011) is a Text Analytics/Natural Language Processing API that offers entity recognition/linking, relation/property extraction, automatic categorization and topic tagging. This API supports entity recognition, disambiguation, linking and topic detection for texts in English, Dutch, French, German, Italian, Polish, Portuguese, Russian, Spanish, and Swedish. TextRazor’s peculiarity consists in linking entities to DBPedia and Wikipedia.
6) Aylien (Dublin, Ireland) has been also functioning since 2011. Its API can boast sentiment analysis (both aspect-based and document-level), entity and concept extraction, document summarization, classification by taxonomy (using pre-defined taxonomies), semantic labeling, language detection, text extraction, related phrases suggestion, etc. Furthermore, Aylien API supports English, French, German, Italian, Portuguese, and Spanish.
7) Another company’s API worth mentioning is Semantria Lexalytics (Boston, USA). The organization launched a cloud text-analyzing API in 2012. Its key text analysis features involve entity extraction, categorization, summarization, sentiment analysis, etc. The API supports 22 languages, including Arabic, Japanese, Chinese, and Turkish.
8) The top-5 is closed by Indico (Boston, USA). This startup was established in 2013, coming out of TechStars Boston. They call themselves a group of innovators whose aim is to revolutionize software through machine learning. Indico Text Analysis comprises sentiment analysis at the document level (demo supports only English), topic detection, keyword extraction, named entity recognition.
9) Haven on Demand is a service developed by Hewlett Packard Enterprise in 2014. It supplies a number of APIs, including APIs for text analysis, i.e. document classification, concept and entity extraction, document categorization, language identification, sentiment analysis, etc.
10) Meaning Cloud is a service launched by a Spanish company Daedalus in February, 2015; it’s a continuation of the previous company’s product – Textalytics. It supports text analysing in 5 languages: English, Spanish, French, Portuguese, and Italian.
Conclusion
As we can see, linguistic APIs by Google’s rivals have no serious reasons to worry, as the corporation has just presented the product first version that doesn’t differ much against the background of other technologies.
All in all, users are those who win in this case. At their disposal there are various competing products, so they can relish affordable prices and high-quality tools.
This table highlights semantic APIs possibilities by different companies and features their prices.
Co-founder - ParallelDots | Unlocking Shelf KPIs for CPGs and Retailers
6yYou might want to add paralleldots.com to your list!
Independent Researcher at n/a - between jobs - who wants me? I want to work!
7yThanks for overview. IMHO, the structure of language is as yet not taken into account sufficiently in building Text Analytics systems. As an example, imagine you want to build a content library of scientific documents on phenomenon A. Sometimes phenomenon A is posited in a tentitative fashion and there is no definite scientific conclusion, sometimes it is present with great certainty. Proper classification depends on this - expressed in English by 'systemic modality': sentences with the word 'may', for instance. . Most classification sysmes do not come close to inforporating thisgrammatical tidbit. I found only one study from Hungary mentioning it. So the race to find the best NLP/API is still on, and very much unfinished. As shameless self-promotion, I am posting a link to a recent paper where I try to take into account some linguistics to improve document classifiaction, and am concluding that it works. I am currently extending my work to MEDLINE. https://www.academia.edu/27207951/Linguistics_improves_statistical_classification_with_KLD_NB_TF_IDF_K-NN_the_positive_effects_of_reducing_feature_dimensionality_or_grammatical_feature_selection._Koos_van_der_Wilt
Conversational Artificial Intelligence Project Leader & Researcher
7ygrazie! interessante!
Enabling Designers, Innovators, and Change Makers with Human Centered Behavioral Science || Grow Products, Innovate Services and Improve Experiences || Consulting, Courses and Tools
7yWould be great if there were better investments in parsing tools... the ones out there are not bad, but far from great
Founder & CEO Regrello (a16z, Tiger, Mubadala)| Advisor McKinsey | Author | ex-CTO Demandbase
7yWell also dissapointing when we ran a benchmark between Google, alchemy Api, we found that it's almost the same Precision recall on real world data. I was expecting something radically better