1
00:00:01,210 --> 00:00:05,260
Embeddings are the core of building LLMs applications.

2
00:00:05,770 --> 00:00:08,260
Let's do a short recap of Embeddings.

3
00:00:09,170 --> 00:00:14,300
Text Embeddings are numeric
representations of text and are used in

4
00:00:14,310 --> 00:00:17,520
actual language processing and machine
learning tasks.

5
00:00:18,550 --> 00:00:23,400
Text Embeddings can be used to measure
the relatedness and similarity between

6
00:00:23,410 --> 00:00:24,940
two pieces of text.

7
00:00:25,550 --> 00:00:28,740
Relatedness is a measure of how closely

8
00:00:28,750 --> 00:00:31,980
two pieces of text are related in meaning.

9
00:00:33,130 --> 00:00:36,420
In this context, the distance between two

10
00:00:36,430 --> 00:00:42,440
Embeddings or two vectors measures their
relatedness which translates to the

11
00:00:42,450 --> 00:00:46,140
relatedness between the text concepts
they represent.

12
00:00:46,870 --> 00:00:49,400
Text concepts are words and phrases.

13
00:00:50,290 --> 00:00:53,820
Similar Embeddings or vectors represent

14
00:00:53,830 --> 00:00:55,100
similar concepts.

15
00:00:56,090 --> 00:00:58,140
There are two common approaches to

16
00:00:58,150 --> 00:01:01,940
measuring relatedness and similarity
between Text Embeddings.

17
00:01:02,610 --> 00:01:06,020
Cosine similarity and Euclidean distance.

18
00:01:06,650 --> 00:01:09,720
We'll find these often used when working

19
00:01:09,730 --> 00:01:10,820
with Embeddings.

20
00:01:11,230 --> 00:01:14,400
Some examples of how Text Embeddings can

21
00:01:14,410 --> 00:01:20,320
be used to measure relatedness and
similarity are text classification which

22
00:01:20,330 --> 00:01:25,460
is the task of assigning a label to a
piece of text, text clustering which is

23
00:01:25,470 --> 00:01:30,700
the task of grouping together pieces of
text that are similar in meaning, and

24
00:01:30,710 --> 00:01:36,840
question answering which is the task of
answering a question posed in natural language.

25
00:01:38,110 --> 00:01:39,740
We'll now take a quick break.

26
00:01:39,990 --> 00:01:43,700
In the next video we'll discuss vector databases.

27
00:01:44,410 --> 00:01:50,840
These databases are designed for storing
and querying high dimensional vectors efficiently.