Ironically but not conscious? This neural network helps you

On the Internet, there is a danger of communicating in words. When others mock you, you will not realize it. On the other hand, sometimes you are misunderstood by the other party. For example, sometimes when someone says "awesome" with a comment, you don't know if he really thinks this thing is good or irony.

Recently, researchers from the University of Lisbon, Spain, cited a new type of deep neural network, CUE-CNN, that can automatically identify sarcasmistic remarks on social media. This will not only avoid the shortcomings of daily text conversations, but it is especially important for natural language processing in the field of artificial intelligence to know whether a particular sentence is ironic.

But this is not a simple question. Generally speaking, computers need to understand whether a sentence is ironic or not. It needs to combine the context of context and find information from the text itself, but it will encounter the problem of insufficient context. After all, words that are exactly the same, including punctuation marks, are completely different when they are spoken by different people. For example, in the same sentence “Make America great again,” which is Trump’s campaign slogan, Trump’s supporters and opponents wrote the sentences on Twitter separately. The meaning would be completely different.

Silvio Amir, a researcher at the University of Lisbon, said: "To understand irony, text information alone is not enough." So, there is a key factor is to understand the user who made the remark .

The team's paper describes a method, is to allow the neural network to find the "users to embed" (User Embeddings). To understand this word, we need to contrast the term "Word Embeddings" that is more common in natural language processing.

There may be near or far contact between words. For example, the words "man" and "king" often appear together and are closely related. "Woman" and "Queen" have similar relationships. The relationship between such words is called "word embedding", and researchers often use mathematical vector space diagrams to represent the relationship. In vector space, words with similar meaning occupy the same position. More intuitive.

So here "user embedding" refers to the relationship between users. Several users may have similar hobbies and similar educational experiences. The relationship between the users of this wave is closer and they will gather in the same position in the space vector.

For example, the paper uses political inclinations as examples. As shown below, each dot represents a user on Twitter: The blue user has at least looked at Democratic politicians: Obama, Hilary, and Bernie Sanders; the red user is following at least one of the Republican politicians: Marco. Rubio, Ted Cruz, Trump. At the same time, attention was paid to the exclusion of both parties' users. From the figure, it is found that users with similar political tendencies gather in the same position in the vector space.

If a blue user makes a remark that supports Trump, which is not the same as other users who are in the same position as him, then the system will determine that this remark is likely to be ironic.

Using a similar approach, in the ideal case one can find a group with a clearly defined boundary. Combining this neural network with contextual clues, the experimental results show that the accuracy rate is 87% and the accuracy rate is 85% compared with other systems. Although accuracy has not improved significantly, considering that this system requires much less manual configuration and monitoring than in the past, it will make its use on social media networks much easier.

Via TC

Handwheel Encoder

The handheld addresser is used to program the address of the monitoring module offline. When in use, connect the two output wires of the handheld encoder to the communication bus terminal (terminal label 1, 2) of the monitoring module, turn on the black power switch on the right side upwards, and press "ten Add", [Subtract ten", [Add one place" and [Subtract one place" to program the address.

Motion Sensor Photocell Light Control,Sensor For Motion Control,Exterior Motion Sensor Control,Hollow Shaft Encoder

Changchun Guangxing Sensing Technology Co.LTD , https://www.gx-encoder.com