Category Archives: Customization

This category describes MeaningCloud’s cutomization engine.

Text Classification in Excel: build your own model

In the previous tutorial we published about Text Classification and MeaningCloud’s Excel add-in, we showed you step by step how to carry out an automatic text classification using an example spreadsheet.

In this tutorial, we are going a bit further: instead of just using one of the predefined classification models we provide, we are going to create our own model using the model customization console in order to classify according to whichever categories we want.

We are going to work with the same example as before: London restaurants reviews extracted from Yelp. We will use some data from the previous tutorial, but for this one we need more texts, so we’ve added some. You can download the spreadsheet here if you want to follow the tutorial along.

If you followed the previous tutorial, you might remember that we tried to use the IAB model (a predefined model for contextual advertisement) to classify the different restaurant reviews and find out what type of restaurants they were. We had limited success: we did obtain a restaurant type for some of them, but for the rest we just got a general category, “Food & Drink“, which didn’t tell us anything new.

This is where our customization tools come in. Our classification models customization console allows you to create a model with the categories you want and lets you define exactly the criteria to use in the classification.

So how do we create this user model?
Continue reading


Learn to develop custom text classifiers (recorded webinar)

Last October 5th we presented our webinar “Learn to develop custom text classifiers with MeaningCloud”. Thank you all for your attendance.

We began by presenting how to do text classification with MeaningCloud and why it is necessary to develop models that are adapted to each specific application scenario. The bulk of the presentation consisted in using a practical case (analysis of restaurant reviews) to show how these models can be developed using our product.

Continue reading


Learn to develop custom text classifiers with MeaningCloud (webinar)

Learn in this webinar how to use MeaningCloud’s tools to create classification models completely adapted to your scenario

Users frequently ask us through our support line how to perform text classification according to application-specific taxonomies. For example, somebody needing to analyzing a bank’s contact center calls and open survey responses might be interested in classifying such messages according to the institution’s different types of products and services (deposits, loans, mortgages, etc.) or the type of interaction (request for information, contracting, complaint, etc.).

Custom classification

Continue reading


Sentiment Analysis in Excel: optimizing for your domain

In previous tutorials about Sentiment Analysis and our Excel add-in, we showed you step by step how to carry out a sentiment analysis with an example spreadsheet. In the first tutorial we focused in how to do the analysis, and then we took a look at the global polarity we obtained. In the second tutorial, we showed you how to customize the aspect-based sentiment analysis to detect exactly what you want in a text through the use of user dictionaries.

In this tutorial we are going to show you how to adapt the sentiment analysis to your own subdomain using of our brand new sentiment model customization functionality.

We are going to continue to use the same example as in the previous tutorials, as well as refer to some of the concepts we explain there, so we recommend to check them out beforehand, specially if you are new to our Excel add-in. You can download here the Excel spreadsheet with the data we are going to use.

The data we have been working on are restaurant reviews extracted from Yelp, more specifically reviews on Japanese restaurants in London.

In the last tutorial, we saw that some of the results we obtained could be improved. The issue in these cases was that certain expressions do not have the same polarity when we are talking about food or a restaurant than when we are using them in a general context. A clear example of this is the verb ‘share’. It is generally considered something positive, but in restaurant reviews it’s mostly mentioned when people order food to share, which has little to do with the sentiment expressed in the review.

This is where the sentiment model customization functionality helps us: it allows us to add our own criteria to the sentiment analysis.

Let’s see how to do this!
Continue reading


Sentiment Analysis in Excel: customizing aspect-based analyses

In the previous tutorial we published about Sentiment Analysis and MeaningCloud’s Excel add-in, we showed you step by step how to do a sentiment analysis using an example spreadsheet. Then we showed you a possible analysis you could obtain with its global polarity results.

In this tutorial we are going a bit further: instead of analyzing the global polarity obtained for different texts, we are going to focus on the analysis of different aspects that appear in them and how to use our dictionaries customization console to improve them and to extract easily the exact information you are interested in.

We are going to work withe same example as before: reviews for Japanese restaurants in London extracted from Yelp. If you don’t have it already from the previous tutorial, you can download the spreadsheet with the data here.

If you followed the previous tutorial, you will remember that when you run the sentiment analysis without changing its default settings, two new sheets appear: Global Sentiment Analysis and Topics Sentiment Analysis. Topics Sentiment Analysis shows you the concepts and entities detected in each one of the texts and the sentiment analysis associated to each one of them.

But what can we do when these are not the aspects of the text we are interested in analyzing? This is where our customization tools come in. Our dictionaries customization console allows you to create a dictionary with any of the concepts or entities you want to detect in your analysis, down the type you want them to have associated.

So how do we create this user dictionary?
Continue reading


A sentiment analysis entirely tailored to your needs with our new customization tool

The adaptation to the domain is what makes the difference between a good sentiment analysis and an exceptional one. Until now, the possibilities of adapting MeaningCloud’s sentiment analysis to your domain relied on the use of personal dictionaries – to create new entities and concepts that the Sentiment Analysis API employed to carry out its aspect-based analysis – or you had to ask our Professional Services Department to develop a tailor-made sentiment model.

Sentiment Models buttonWith the release of Sentiment Analysis 2.1, we incorporated a new customization tool designed to facilitate the creation of personal sentiment models. This tool fully employs our Natural Language Processing technology to enable you to be autonomous and develop —without programming— powerful sentiment analysis engines tailored to your needs.

Other tools for customizing sentiment analysis available on the market, mostly permit to define “bags of words” with either positive or negative polarity. Our tools go far beyond and enable you to:

  • Define the role of a word as a polarity vector (container, negator, modifier), allowing to use lemmas to easily incorporate all the possible variants of each word
  • Specify particular cases of a word’s polarity, depending on the context in which it appears or its morphosyntactic function in each case
  • Define multiword expressions as priority elements in the evaluation of polarity
  • Manage how these custom polarity models complement or replace the general dictionaries of every language.

Screenshot Sentiment Customization

For example, the expression “the interest rate is very high” expressed by a financial service customer may be positive if it refers to deposits, but negative if it has to do with mortgages. With this tool, it is possible to define these different polarities for each case.

And, the use of this tool is included in your MeaningCloud subscription at no additional cost (even in the Free plan).

This sentiment models tool complements our offer for the development of custom semantic resources and contributes to the goal of MeaningCloud of making the highest-quality text analytics available to all developers.

Would you like to know how to apply the sentiment analysis customization tool in a practical scenario? Register for this webinar on May 4th and you will find out.

UPDATE: This webinar has already taken place. See the recording here.

IMPORTANT: Sentiment Analysis 2.1 introduces changes to the API that make it necessary to migrate your applications to this new version. Migration is very simple, and it is explained here. Remember that Sentiment Analysis 2.0 will no longer be operating as of July 7, 2016: plan your migration with time!

Feature-level sentiment analysis

Back when we were called Textalytics, we published a tutorial that showed how to carry out feature-level sentiment analysis for a specific domain: comic book reviews.

Cover for Marvel's Black Widow #1

Marvel’s Black Widow #1

Since then, besides changing our name, we have improved our Sentiment Analysis API and how to customize the different analyses through our customization engine. In this post we are going to show you how to do a feature-level sentiment analysis using MeaningCloud.

One of the main changes in the latest release of our API is the possibility of using custom dictionaries in the detailed sentiment analysis provided by the Sentiment Analysis API. We are going to use comic book reviews to illustrate how to work, but the same process applies to any other fields where sentiment comes into play, such as hotel reviews, Foursquare tips, Facebook status updates or tweets about a specific event.

Continue reading


Exploring Social Media for Healthcare Data

People enjoy sharing information through social media, including healthcare data. Yeah, it is true! And it constitutes the starting point of the research work titled ‘Exploring Spanish health social media for detecting drug effects’, which aims at following social media conversations to identify how people talk about their relation with drug consumption. This allows identifying possible adverse effects previously unknown related to these drugs. Although there is a protocol to communicate to the authorities the identification of a drug adverse effect, only a 5 – 20% of them are reported. Besides, conversations around drugs, symptoms, conditions and diseases can be analyzed to learn more about them. For example, it is possible to see how people search for specific drugs using social media, while others sell them, perhaps illegally. Many others talk about mixing alcohol with drugs or other illegal substances. Of course, one cannot believe everything that appears on the Internet this is another issue—, but it can highlight some hypothesis for further research.

drugs

Some researchers from the Advanced Databases Group at Carlos III University of Madrid have carried out the mentioned study, designing hybrid models to capture the needed knowledge to identify adverse effects. The Natural Language Processing platform which supports the implementation of the analysis process based on such models is MeaningCloud. The customization capabilities provided by the platform have been decisive to include specific vocabulary and medical domain knowledge. As we know, the names of drugs and symptoms might be complex and, in some cases, difficult to write properly. The algorithm’s results are promising, with a 10% increase in recall when compared to other known algorithms. You can find further details in the scientific paper published by the BMC Medical Informatics and Decision Making Journal.

These developments have been part of the TrendMiner project, and are now available in the prototype website TrendMiner Health Analytics Dashboard, which shows people’s comments about antidepressants gathered from social media. The console displays the mentions of antidepressants and related symptoms and, by clicking on any of them, their evolution over time. Moreover, the source texts analyzed to compute those mentions are shown at the bottom, with labels highlighting the names of drugs, symptoms or diseases, and any relations among them. Such relations might say if a drug is indicated for a symptom or if a disease is an adverse effect of the mentioned drug. The prototype also allows searching by the ATC code (Anatomical Therapeutic Chemical Classification System) and the corresponding level according to this classification scheme. So, if you mark the ‘By Active Substance’ selector, you are searching any drug containing the active substance of the product you inserted in the search box. Furthermore, the predictive search functionality makes easier to find the right expression for a drug or disease. Please, have a look at the prototype and tell us what you think about it. If you find a chart useful, you can even tweet it from there! Any comment is more than welcome.


Sentiment Analysis 2.0: Migration guide

We have released a new version of one of our more popular APIs: Sentiment Analysis. In Sentiment Analysis 2.0:

  • The rules used for defining polarity terms have been greatly improved, adding new operators and making the models used much more flexible, which in turn leads to better results.
  • Sentiment analysis is now done at more levels, allowing to identify more complex syntactic structures and to obtain more detailed information about how the polarity is expressed.
  • More configuration options have been added related to the morphosyntactic analysis over which the sentiment analysis is carried out.
  • The architecture of the service has changed, leading to a tenfold improvement in the response time.
  • An integration with the Lemmatization, PoS and Parsing API has been added in order to ease the way of creating applications that use the information provided by both APIs.
  • Dictionary customization has been fully integrated in order to get out the most out of its functionality.

All these improvements mean the migration process is not as fast as it would be with a minor version. These are the things you need to know to migrate your applications from Sentiment Analysis 1.2 to Sentiment Analysis 2.0.
Continue reading


Customize your text analytics tools (recorded webinar)

Last May 14th we presented our “Boost your text analytics accuracy” webinar.

We discussed why we need to customize text analytics processes -by including domain-based information- to improve the accuracy (precision, recall) of these tools. And we did a walk-through of MeaningCloud’s features to customize several of its functions:

  • Text classification
  • Information extraction
  • Sentiment analysis.

These customization tools feature graphical user interfaces and are very easy to use, thus empowering the users to adapt the system to their applicactions and putting high-quality text analytics at everybody’s fingertips. We are confident that these features, together, are unique in the industry and put MeaningCloud ahead of the competitors’ offerings.

For those of you interested, below you can find the webinar’s slides and recording.

(También presentamos este webinar en español. Tenéis la grabación aquí.)

MeaningCloud Webinar – Better Text Analytics Using Customization Tools from MeaningCloud on Vimeo.