We have just published a new release of MeaningCloud that affects Topics Extraction, Lemmatization, POS and Parsing, and Text Classification APIs. Although there are several new features in terms of new functionalities and parameters, the most important aspect of this release lies under the hood and essentially consists of a refactoring of the way in which concept-type topics are internally handled, much more in line with the use of other semantic resources. This lays the foundations for better performance and new features related to the extraction of this type of information. Sty tuned for great improvements in this area in future releases.
The other two great lines of this release are the enrichment of the morphosyntactic analysis with information extraction and sentiment analysis elements (which enable new and richer types of analyses that combine the text’s structure with topics and polarity) and a new predefined classification model.
Here are some details about the developments in the different APIs:
Topics Extraction 2.0
The main new features in this API consist of internal improvements that will increase the quality of the extraction of concepts. But there are other relevant aspects, such as:
- Greater flexibility in the use and traceability of user dictionaries.
- The new ilang parameter (interface language) allows to specify the language in which results are returned. This is useful when it comes to extracting topics that have multiple forms expressed in different languages (for example, city names), as it permits to select a default form.
- A new type of topic, quantity expressions, which enables to extract quantities both as absolute values (for example, meters, liters, etc.) and percentages.
- Various improvements in the use of parameters and the presentation of results, in order to make them easier and more homogeneous.
Lemmatization, POS and Parsing 2.0
As in the case of the previous API, the main new features of this one consist of internal improvements that will increase the quality of the extraction of concepts. At the same time, Lemmatization, POS and Parsing 2.0 incorporates other substantial developments:
- This API is still the morphosyntactic and semantic analysis engine of MeaningCloud, but now its functionality has been extended with topics extraction and sentiment analysis. In this new version, it is possible to get the topics and the sentiment associated with each structural element of the text. This allows much more accurate analyses that combine the structure of the text with the various topics and opinions appearing on it; for example, it is possible to carry out a sentiment analysis centered on a brand (affinity) and differentiate users who own the product from those who do not.
- By including information extraction functionalities, the new release of this API incorporates the new developments of Topics Extraction regarding the use and the traceability of the user dictionaries, the ilang parameter and the new type of topic, quantity expressions.
- In addition, Lemmatization, POS and Parsing 2.0 incorporates several improvements in the use of parameters and in the presentation of results, in order to increase uniformity and simplicity.
Text Classification 1.2
The catalog of predefined Text Classification models is growing with a new addition: Social Media. It is an optimized model to understand conversations and posts in all kinds of social media (Twitter, Facebook, forums…) using a simple taxonomy of 17 categories ranging from Economy to Sports, Greetings and Thanks. The classification with Social Media is the easiest way to understand what a social comment is about. This model is available both in Spanish and in English for all Text Classification users. More information here.
IMPORTANT: the changes dealing with the parameters and results of the APIs require you to migrate your applications to this new release. The current versions of Topics Extraction, and Lemmatization, POS and Parsing will cease to be operative as of February 29th, 2016, so you can plan your migration in time. In the previous posts, you can access detailed guides (for Topics Extraction and Lemmatization, POS and Parsing) that explain thoroughly the changes to make and help you in the migration process.
We hope you will benefit from this new release.