The basis of text sentiment analysis is natural language processing, affective dictionary, machine learning method and so on. Here are some of the resources I've summed up.

The basis of text sentiment analysis is natural language processing, affective dictionary, machine learning method and so on. Here are some of the resources I've summed up.

Dictionary resources:
Sentiwordnet
"Knowledge Network" Chinese version
Chinese Affective polarity dictionary NTUSD
Emotion Vocabulary Ontology Download

Natural language processing tools and platforms:
Institute of Social Computing and Information retrieval, Harbin Institute of Technology
ISNOWFY/SNOWNLP GitHub

Chinese participle:
Natural language processing and information retrieval sharing platform nlpir.org
Fxsjy/jieba GitHub

Corpus Resources:
Information Classification and Emotion discovery

Course:
The seventh course of natural language processing, Stanford University, "Affective Analysis (sentiment analyses)"

Websites and blogs:
Text Classification for sentiment analysis
Second try:sentiment analysis in Python:andy Bromberg
Basic sentiment analysis with Python

Paper:
HTTP/citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.244.9480&rep=rep1&type=pdf

Tools:
It is recommended to use Python. Integration of all the above features, easy to learn.

I wrote a blog, basically have a simple implementation of the above steps, thick-skinned stickers.
Explore in Data

The current situation from the three-block application of the topic, near-field speech and facial recognition almost can be industrialized, at present a large number of mature technology applications, the basic can meet the application needs, of course, these two pieces of technology is also involved in some special scenes of technical coverage is not enough, for example, voice this piece of far-field program is very poor , the facial recognition of this piece if the time span long scene, the effect is very poor. (such as in childhood and in contrast to the present), it is undeniable that the standard of industrial application has been achieved in some basic application scenarios. From the current open API two, basically still take this technology development when the big platform development opportunities to see, but at present the application level is not good product follow-up, this piece will be a good opportunity point. Not finished, to be liked over 10 again continued. Thank you for your praise, come back again a wave of re-examination of the question, first answer the questions of the Lord. If the main question is whether the two platforms are reliable, my proposed two (speech recognition and image recognition) can be used in industrial applications, and the speech synthesis module is also basic to ensure the availability. The other includes the user portrait, the recommendation algorithm part at present to the user own data to be very big, considers the platform data security, at present this piece is not necessary, needs to consider in the safeguard data security premise application. NLP and other modules currently from the algorithm capabilities, including specific applications are still many problems, it is recommended to observe a period of time. At present, Baidu Brain external application of speech recognition part of the function is a full highlight. If the main question is Microsoft and Baidu in the algorithm capabilities will be what difference, the current open API capabilities are basically similar, and these platform resources to ensure that the service is relatively stable, and can accommodate a certain amount of throughput, if the main wish in the domestic application, recommended priority to try to provide the ability of Baidu brain side. And finally, there's some extra crap for this kind of competency platform. Currently, the algorithm in the enterprise has the direction of algorithm research and application oriented two directions, the research direction is generally in the form of Enterprise Internal Research institute: such as the IDL of Baidu, another piece to application-oriented, For example, the first-line enterprise's algorithm team or the new organization that is directly based on artificial intelligence, this kind of structure usually starts with the artificial intelligence application for certain scene, carries on the algorithm accumulation and the application, the innovation and the research direction is weaker than the institute. and the basic ability of open platform is generally by the Institute of such organizations, the function of the current can be applied mainly to the existing product capacity expansion, other applications such as machine learning, map construction, recommendations and so on due to data security and application scenarios, the simple open form of the algorithm also needs market verification. In addition, this kind of algorithm + resources open platform itself shows that the next stage in the application level algorithm, computing resources has become increasingly not the bottleneck of the application of artificial intelligence technology, the first wave of application innovation platform for the industry, the opportunity is imminent, and then the new opportunity will be gradually overcome with the algorithm, The breakthrough in every technology in the field of artificial intelligence means breakthrough product changes in a number of areas.At present, there are many enterprises in the direction of finance and hardware, the intelligent technology application actually has many applications in the backend, but the products in front-end packaging have many opportunities in the direction of entertainment, social, game and so on.

The basis of text sentiment analysis is natural language processing, affective dictionary, machine learning method and so on. Here are some of the resources I've summed up.