AI Note 3: How to use WPS AI to process Excel data and improve data processing efficiency

I. Introduction

WPS Win client version: 11.1.0.14650

In the past few days, I discovered that WPS's light form has quietly launched WPS AI.
There are currently 5 supported functions, namely: formula writing, conditional formatting, intelligent classification, intelligent extraction, and sentiment analysis.

Let's test these functions in turn.

2. Functional test

One of the scenarios used in this test is retail goods. The data used is mainly generated by ChatGPT, and some of them are modified by ourselves.
Two worksheets are involved:

• Purchase and evaluation record table: a total of 30 order data, 8 columns, the column names are order number, store name, product ID, price, quantity, user ID, purchase time, and evaluation. The specific data are as follows:
• Commodity description table: 5 commodities in total, 3 columns, and the column names are commodity ID, commodity name, and commodity description.
  

2.1 Write the formula

Writing formulas is the highlight, which is a great essence of Excel! Returns a formula that can be used to insert into a cell.
This time we mainly test several commonly used formulas: sum (sum), condition (if), and retrieval (vlookup).

Let’s take a look at the total number of shoes sold. The demand is transformed into [SUM D column], and then AI gives an answer [=SUM(D2:D31)], which can be directly calculated by inserting it into the cell The corresponding value is not bad!

Prompt: sum column D

This is just a method. After actual testing, you can also directly specify the corresponding column names. For example, the above requirement can be converted into [summing the number of columns], and the result is the same:

Prompt: sum the amount column

Next, try conditional judgment, you need to add a new column to operate. Classify and label users. When the purchase amount is greater than 500, it is marked as a type A user, and the others are marked as B type users.
When writing the prompt rules, I found that I missed summing the D column. After sending it, I thought it had to be changed. I didn’t expect AI to automatically add it. It’s a little surprise!

Prompt: Calculate the sum of column D according to the grouping of column F. If the result value is greater than 500, return "type A", otherwise return "type B"

Next, use vlookup to match the product name. good! one time~~

Prompt: match the product ID in the current worksheet with the product ID in the product description table, and finally return the product name in the product description table

To increase the difficulty, it can be combined with column name retrieval to batch match the values ​​in the product table. For example, return the value of the [Product Name] column in the [Product Description Table] in the [Product Name] column, and return the value of the [Product Description] column in the [Product Description Table] in the [Product Description] column.
In fact, it is essentially to write two vlookups to match [product name] and [product description], and then combine these two into one formula so that you can drag and drop to achieve batch matching.
I tried to write several Prompts, but they couldn't be realized, so I posted two of them here.

Prompt: match the product ID in the current worksheet with the product ID in the product description table, return the value of the product name column in the product description table in the product name column of the current worksheet, and return the product description table in the product description column of the current worksheet The value of the product description column in
the return result: =VLOOKUP(C2,commodity description table!A:B,2,FALSE)

Prompt: match the product ID in the current worksheet with the product ID in the product description table, return the value of the product name column in the product description table in column D, and return the value of the product description column in the product description table in column E Return results
: Sorry, the formula cannot be generated according to your current needs, please change the description and try again

I don't know if AI can realize such a function. From the current tips, it can't be realized. I don't know if there are better tips.
The reference formula of this logic is as follows, and it can be matched to the corresponding column value of [Product Description Table] when placed in column D and E.

Formula: =VLOOKUP( $C2,Commoditydescription!$ A:$C,MATCH(E$1$$1,Commoditydescriptiontable!$ A$1:$ C$1.0),FALSE)

2.2 Conditional formatting

Conditional formatting is to specify some rules and then filter out the corresponding cells. For example: identify and color duplicate values, color cells greater than a certain value, etc.
The returned data acts on the original column.

In the following case, after entering Prompt, you can see a "thinking chain" process returned by AI. First, define a formula to determine the top 5, and then add conditional formatting to column F and set it to green.

Prompt: Mark the top 5 cells in the amount column as green

If you are not satisfied with the color, you can also specify the hexadecimal color number yourself, the example is as follows

Prompt: Mark the top 5 cells in the amount column as color: 00B050

Mark duplicate values ​​in red, this is similar to the above, nothing special.

Prompt: Mark duplicate values ​​in column H in red

2.3 Intelligent Classification

It is to classify according to the existing column data, and you need to specify the classification name yourself, which is a bit like clustering in the algorithm.
The returned data is placed in a new column.
Here is an example of a short comment, the steps are as follows:

• In the function tab, select [Smart Classification]
• Select [column L-short comment] in the field, and then send
• Enter the category in the suggested card. There are three categories in short reviews: positive reviews, neutral reviews, and negative reviews. I made it difficult for some AI to only give two categories, namely evaluation category 1 and evaluation category 2, and see how it divides.
• Click Execute Classification and wait for the classification to complete

From the screenshots, we can see that the result of intelligent classification is to divide the neutral and negative reviews into the same category. Is it either positive or something else, hahaha.

2.4 Intelligent extraction

Smart extraction is to extract some substrings in the string.
The returned data is placed in a new column.

Let's take the evaluation to extract, the steps are as follows:

• Select the function tab [Smart Extraction]
• Select [Column K-Evaluation] in the field, and then send
• Enter the words to be extracted in the prompt card. Here, the test extracts 3 words, which are fashion, comfort, and poor.
• Click Execute Extraction and wait for the execution to complete.

The execution result returns three columns, one for each keyword.
The process and result diagrams are as follows:

The returned result is not necessarily the specified word, and it seems that a layer of processing will be performed. For example, in the figure below, the fashion column also contains some words that do not contain fashion. Part of it seems like a mystery, [bad leather and good leather] are also fashionable?

2.5 Sentiment Analysis

Sentiment analysis should be through the analysis of natural language, and then judge the emotional color of the entry, and identify whether the emotion of the text is positive, neutral, or negative.
The returned data is placed in a new column.

Let's take the evaluation to do sentiment analysis, the steps are as follows:

• Select the function tab [Sentiment Analysis]
• Select [Column K-Evaluation] in the field, then send it, and wait for the execution to complete.

The execution result returns a column: sentiment analysis, scoring and calculating positive, neutral, and negative according to the color of the text.

Judging from the results, some are unreliable, three [default praise], two neutral, and one positive.

3. Summary

According to the current test, writing formulas and conditional formatting is more reliable. The three functions of intelligent classification, intelligent extraction and sentiment analysis seem to be less useful, and the effect does not seem to surprise.
It doesn't matter if you don't understand the formula in the future, as long as you know Prompt, you can mobilize AI to help.
For particularly complex formulas, they can actually be split into many sub-formulas and then spliced ​​together, so the understanding and disassembly of requirements are also relatively important. With the help of AI, basically more than 90% of the scenarios should be able to easily solve.
What remains may be the frontier of AI, and it's not yet clear where that frontier is.

#Past Review

WPS-AI Experience Note 1: Generate PPT with one click

WPS-AI Experience Note 0: Obtain "Admission Ticket" and "Ready to Enter"

- End -