Sentiment Analysis - Huawei FusionInsight

Sentiment Analysis Engine
Function Highlights

Latest APIs

No tags required

High accuracy

Providing additional semantic information

Functional View
Function Description

Technology

The LSTM model builds models based on the comment data of the last month in Amazon. The newly built model is trained and then used to encode input text, for example, encoding 4096 components as a one-dimensional vector. The sentiment value ranges from -4 to 2, the negative sentiment value ranges from -4 to 0, and the positive sentiment value ranges from 0 to 2. The percentage of negative sentiment in comments is greater than that of positive sentiment. The feature-based model is flexible and can process misspelled words, local language, and tags, for example:

sentiment(“teribl”) = -0.41132

sentiment(“wtf”) = -0.424688

sentiment(“:)”) = 0.396196

GUI

Definition

Input

{

“text”: &ltstring>

}

text: The input text for sentiment_score. Repeat this parameter to obtain scores for multiple text inputs.

Output:

{

“error”: {

“code”: &ltinteger>

“message”: &ltstring>

“data”: {

sentiment_score”: &ltfloat>

}

}

error: optional (present when there is an error)

data: optional (present when the call is successful) which will have the sentiment_score of the supplied texts

Example

Input

{

“text”: “It was a wonderful day”,

“text”: “It was a normalday”

}

Output:

{

“data”: {

“sentiment_score”: 0.5034,

“sentiment_score”: 0.071

}

}