> 2021年07月30日信息消化 ### How to Read a Book at Higher Order Reading Skill Level origin: [How to Read a Book at Higher Order Reading Skill Level](https://medium.com/personal-growth/how-to-read-a-book-at-higher-order-reading-skill-level-1347e59cc681) > Total immersion is how you learn from great authors I read on my phone, swipe through pages on my iPad. And read physical a book in the morning just before deep work and in the evening before bed. I also read whenever I get the chance in the day. Over 90 percent of the apps on my phone are for reading. Seneca, a Roman Stoic philosopher, once said, “You must linger among a limited number of master-thinkers, and digest their works if you would derive ideas which shall win firm hold in your mind.” ### 5 Mental Shifts That Helped Me Build a 6-Figure Online Business in My 20s origin: [5 Mental Shifts That Helped Me Build a 6-Figure Online Business in My 20s](https://medium.com/the-post-grad-survival-guide/5-mental-shifts-that-helped-me-build-a-6-figure-online-business-in-my-20s-47d1b2e15de4) All I knew is that I wanted to be independent — financially, locally, and mentally. I wanted to work whenever, wherever I want and be able to choose the projects I work on. Today, four years later, I’m in that fortunate place where I’m living that reality: I built a 6-figure online business [with my better half](https://entrepreneurshandbook.co/how-to-successfully-run-a-business-with-your-better-half-d4e345eaf7d0?source=search_post---------0), work on projects that excite me, and experience the freedom I dreamt of. But whenever I tell my story, a huge part is missing because most people only see the [tip of the iceberg](https://medium.com/publishous/to-succeed-you-need-to-stop-looking-at-the-tip-of-the-iceberg-7e671e98a353). They see my current numbers, the current reality, and the freedom I experience right now. ##### From Reactive to Decisive You’ll never find satisfaction in business if you act as if you’re the puppet of your customers. Being flexible and adapting to changes is important, but so is staying true to your core principles and charging the fee you want or need. Everybody wants discounts and special rates, but you'll end up starving if you say yes and amen to everyone. As an entrepreneur, you need to know your worth and then add tax. **A lot** of tax. ##### You Can’t Always Do What You Love If you want to build a successful business, you need to put your ego aside and do stuff that sucks until you can outsource it. You’ll need to learn new skills, relearn things, and delegate tasks to people who can do it more efficiently than yourself. Even the simplest business model will require you to do *some* things you won’t enjoy. You need to love the whole journey of entrepreneurship, including all the ups and downs. ##### You Don’t Need to “Hustle More” You need both: You need to believe in yourself and you need to work your ass off. But you also need ***so much more\***. If you don’t believe that you can win, you won’t *act* like a winner. And if you don’t do the work, you won’t see the results. But business success comes down to optimizing your processes, being efficient, delegating tasks, and increasing the time you spend on [*high-value activities*](https://www.productiveflourishing.com/what-are-your-high-value-activities/). Sometimes, putting in the work and hustling more can be the solution to a more successful business. But in most cases, it’s not about how hard you work but about how *smart* your processes and structures are. As an entrepreneur, you need to ensure that you don’t waste your time on redundant or repetitive tasks. Otherwise, you‘ll likely find yourself in a rat race that you created for yourself. The earlier you start creating a blueprint and well-thought structures for certain business activities, the better. ##### Being Fast Is Better Than Being Perfect At the beginning of my entrepreneurial journey, I tried to be everybody’s darling and do everything *perfectly.* I spent hours designing PowerPoint slides, invoices, or business cards. And I felt good because I thought that I was doing important work for my business. But the truth is that I was just wasting time. In most cases, perfectionism is nothing more than a roadblock and getting it done is better than trying to be perfect. If you’re always trying to be *perfect*, you’ll likely end up being slow. And being slow can quickly turn into a disadvantage if you’re not the only one in your market. Being fast and adapting to changes quickly, on the other hand, can be a strategic advantage. ### Word2Vec Explained origin: [Word2Vec Explained](https://towardsdatascience.com/word2vec-explained-49c52b4ccb71) ##### What is a Word Embedding? Word embeddings is a technique where individual words are transformed into a numerical representation of the word (a vector). Where each word is mapped to one vector, this vector is then learned in a way which resembles a neural network. The vectors try to capture various characteristics of that word with regard to the overall text. These characteristics can include the semantic relationship of the word, definitions, context, etc. With these numerical representations, you can do many things like identify similarity or dissimilarity between words. ``` For example: have = [1, 0, 0, 0, 0, 0, ... 0] a = [0, 1, 0, 0, 0, 0, ... 0] good = [0, 0, 1, 0, 0, 0, ... 0] day = [0, 0, 0, 1, 0, 0, ... 0] ... ``` ##### Word2Vec Architecture The effectiveness of Word2Vec comes from its ability to group together vectors of similar words. Given a large enough dataset, Word2Vec can make strong estimates about a words meaning based on their occurrences in the text. These estimates yield word associations with other words in the corpus. For example, words like “King” and “Queen” would be very similar with one another. When conducting algebraic operations on word embeddings you can find a close approximation of word similarities. For example, the embedding vector of "king" - the embedding vector of "man" + the embedding vector of "woman" yielded a vector which is very close to the embedding vector of "queen". ``` [5,3] - [2,1] + [3, 2] = [5,4] ``` ##### CBOW (Continuous Bag of Words) This architecture is very similar to a feed forward neural network. This model architecture essentially tries to predict a target word from a list of context words. The intuition behind this model is quite simple, given a phrase `"Have a great day"` , we will chose our target word to be “a” and our context words to be [“have”, “great”, “day”]. What this model will do is take the distributed representations of the context words to try and predict the target word. 这种架构与前馈神经网络非常相似。这个模型架构基本上是试图从一个上下文词列表中预测一个目标词。这个模型背后的直觉很简单，给定一个短语 "Have a great day"，我们将选择我们的目标词为 "a"，我们的上下文词为["have", "great", "day"]。这个模型要做的是利用上下文词的分布式表示来尝试预测目标词。 ##### Continuous Skip-Gram Model The skip-gram model is the literal opposite of the CBOW model. In this architecture, it takes the current word as an input and tries to accurately predict the words before and after this current word. This model essentially tries to learn and predict the context words around the specified input word. Given the sentence `"Here we go again"` , we can choose our input word to be “go” and this model will try to predict [“here”, “we”]. Based on experiments assessing the accuracy of this model it was found that the prediction quality improves given a large range of word vectors, however it also increases the computational complexity. 跳格模型与CBOW模型的字面意思相反。在这个架构中，它把当前的词作为输入，并试图准确地预测这个当前词之前和之后的词。这个模型本质上是试图学习和预测指定输入词周围的上下文词。给出一个句子 "我们又来了"，我们可以选择我们的输入词为 "go"，这个模型将尝试预测["here", "we"] 。根据评估这个模型的准确性的实验发现，在给定一个大范围的词向量的情况下，预测质量有所提高，然而这也增加了计算的复杂性。 ##### Implementation I’ll be showing how to use word2vec to generate word embeddings and use those embeddings for finding similar words and visualization of embeddings through PCA. 我将展示如何使用word2vec来生成单词嵌入，并使用这些嵌入来寻找相似的单词，以及通过PCA对嵌入进行可视化。 ##### Requirements ``` nltk==3.6.1 node2vec==0.4.3 pandas==1.2.4 matplotlib==3.3.4 gensim==4.0.1 scikit-learn=0.24.1 ``` ```python import pandas as pd import nltk import string import matplotlib.pyplot as plt from nltk.corpus import stopwords from nltk import word_tokenize from gensim.models import Word2Vec as w2v from sklearn.decomposition import PCA # constants PATH = 'data/shakespeare.txt' sw = stopwords.words('english') plt.style.use('ggplot') # nltk.download('punkt') # nltk.download('stopwords') # import data lines = [] with open(PATH, 'r') as f: for l in f: lines.append(l) ``` ##### Preprocess Data ```python # remove new lines lines = [line.rstrip('\n') for line in lines] # make all characters lower lines = [line.lower() for line in lines] # remove punctuations from each line lines = [line.translate(str.maketrans('', '', string.punctuation)) for line in lines] # tokenize lines = [word_tokenize(line) for line in lines] def remove_stopwords(lines, sw = sw): ''' The purpose of this function is to remove stopwords from a given array of lines. params: lines (Array / List) : The list of lines you want to remove the stopwords from sw (Set) : The set of stopwords you want to remove example: lines = remove_stopwords(lines = lines, sw = sw) ''' res = [] for line in lines: original = line line = [w for w in line if w not in sw] if len(line) < 1: line = original res.append(line) return res filtered_lines = remove_stopwords(lines = lines, sw = sw) ``` ##### Embed ```python w = w2v( filtered_lines, min_count=3, sg = 1, window=7 ) print(w.wv.most_similar('thou')) emb_df = ( pd.DataFrame( [w.wv.get_vector(str(n)) for n in w.wv.key_to_index], index = w.wv.key_to_index ) ) print(emb_df.shape) emb_df.head() ```  ##### PCA on Embeddings ```python pca = PCA(n_components=2, random_state=7) pca_mdl = pca.fit_transform(emb_df) emb_df_PCA = ( pd.DataFrame( pca_mdl, columns=['x','y'], index = emb_df.index ) ) plt.clf() fig = plt.figure(figsize=(6,4)) plt.scatter( x = emb_df_PCA['x'], y = emb_df_PCA['y'], s = 0.4, color = 'maroon', alpha = 0.5 ) plt.xlabel('PCA-1') plt.ylabel('PCA-2') plt.title('PCA Visualization') plt.plot() ```  Words similar to each other would be placed closer together to one another. Tensorflow has made a very beautiful, intuitive and user friendly representation of the word2vec model. I highly recommend you to explore it as it allows you to interact with the results of word2vec. The link is below. https://projector.tensorflow.org/ ##### Concluding Remarks Word embeddings are an essential part of solving many problems in NLP, it depicts how humans understand language to a machine. Given a large corpus of text, word2vec produces an embedding vector associated to each word in the corpus. These embeddings are structured such that words with similar characteristics are in close proximity of one another. CBOW (continuous bag of words) and the skip-gram model are the two main architectures associated to word2vec. Given an input word, skip-gram will try to predict the words in context to the input whereas the CBOW model will take a variety of words and try to predict the missing one. 词嵌入是解决NLP中许多问题的重要部分，它向机器描述了人类如何理解语言。给定一个大的文本语料库，word2vec产生一个与语料库中每个词相关的嵌入向量。这些嵌入的结构是这样的：具有相似特征的词彼此相近。CBOW（连续词包）和skip-gram模型是与word2vec相关的两个主要架构。给定一个输入词，skip-gram将试图预测与输入词有关的上下文，而CBOW模型将采用各种词并试图预测缺失的词。