我正在使用IBM Watson的自然语言理解API从URL中获取关键字和实体。我想迭代JSON响应以获取所有关键字和实体，并将它们填充到我的results.html文件中。
我试图遍历application.py文件和使用jinja的results.html文件中的结果。
helpers.py文件返回一个json.dump并将其发送到我的application.py文件，以便我可以遍历结果。
但是，我收到以下错误：
TypeError: string indices must be integers
我查了json.dump vs json.load以及字符串和字典来帮助解决这个问题，但我无法让代码工作。如果您需要更多信息，请告诉我。我需要在今年年底之前解决这个问题。先感谢您。
这是我的applications.py文件
@app.route("/URL", methods=["GET", "POST"])
def URL():
"""Analyze URL."""
# if user reached route via POST (as by submitting a form via POST)
if request.method == "POST":
# if nothing was entered return apology
if not request.form.get("URL"):
return apology("please enter a URL")
URL = request.form.get("URL")
# analyze URL using analyze function in helpers.py
results = analyze(request.form.get("URL"))
for item in results:
keywords = item["keywords"]["text"]
entities = item["entities"]["text"]
return render_template("results.html", results=results, URL=URL)
# check if URL is valid
if not results:
return apology("this is not a valid URL")
else:
return render_template("url.html")
这是helpers.py文件。
def analyze(URL):
natural_language_understanding = NaturalLanguageUnderstandingV1(
version='2017-02-27',
username='MUSTGETYOURUSERNAME',
password='MUSTGETYOURPASSWORD')
response = natural_language_understanding.analyze(
url=URL,
features=Features(entities=EntitiesOptions(), keywords=KeywordsOptions()))
return (json.dumps(response, indent=2))
以下是使用jinja的results.html文件：
{% extends "layout.html" %}
{% block title %}
Results
{% endblock %}
{% block main %}
<h2>Powered by IBM Watson's AI to recommend your #'s and @'s
for tweeting</h2>
<p>{{URL}}</p>
{% for item in results %}
<tr>
<td>{{ item.keywords }}</td>
<td>{{ item.entities }}</td>
</tr>
{% endfor %}
<a class="twitter-share-button"
href="https://twitter.com/intent/tweet">Tweet</a>
{% endblock %}
以下是输出结果：
[
{
"text": "Android apps",
"relevance": 0.926516
},
{
"text": "Chrome OS",
"relevance": 0.878045
},
{
"text": "Sorry Android fanboys",
"relevance": 0.696885
},
{
"text": "Android tablet",
"relevance": 0.695471
},
{
"text": "absolutely wonderful Android",
"relevance": 0.672889
},
{
"text": "Chrome OS beta",
"relevance": 0.626619
},
{
"text": "Android Police",
"relevance": 0.592994
},
{
"text": "Chrome OS devices",
"relevance": 0.566831
},
{
"text": "count Android",
"relevance": 0.563911
},
{
"text": "dominant Google OS",
"relevance": 0.553724
},
{
"text": "Chrome Unboxed",
"relevance": 0.540076
},
{
"text": "overall tablet sales",
"relevance": 0.511826
},
{
"text": "inexpensive Google rival",
"relevance": 0.498259
},
{
"text": "half incremental improvements",
"relevance": 0.468663
},
{
"text": "standard operating procedure",
"relevance": 0.45946
},
{
"text": "uncommon Chromebook form",
"relevance": 0.456969
},
{
"text": "content consumption machines",
"relevance": 0.451775
},
{
"text": "absolute best pieces",
"relevance": 0.450763
},
{
"text": "content creation ones",
"relevance": 0.450345
},
{
"text": "rich new fusion",
"relevance": 0.446127
},
{
"text": "Amazon Fire tablet",
"relevance": 0.444685
},
{
"text": "selling tablet",
"relevance": 0.444241
},
{
"text": "tablet operating",
"relevance": 0.440434
},
{
"text": "Google Pixelbook",
"relevance": 0.440007
},
{
"text": "Google store",
"relevance": 0.439719
},
{
"text": "cheap tablets",
"relevance": 0.408395
},
{
"text": "immortal highlander",
"relevance": 0.404233
},
{
"text": "disparate OSes",
"relevance": 0.401626
},
{
"text": "laptop space",
"relevance": 0.40117
},
{
"text": "detachable two-in-one",
"relevance": 0.396257
},
{
"text": "pleasant surprises",
"relevance": 0.394027
},
{
"text": "additional oomph",
"relevance": 0.393127
},
{
"text": "Samsung",
"relevance": 0.391534
},
{
"text": "flashy Chromebook",
"relevance": 0.391359
},
{
"text": "sleek Chromebook",
"relevance": 0.390035
},
{
"text": "smaller devices",
"relevance": 0.389106
},
{
"text": "operating systems",
"relevance": 0.388958
},
{
"text": "new feature",
"relevance": 0.388395
},
{
"text": "true multitasking",
"relevance": 0.388097
},
{
"text": "tablet-like device",
"relevance": 0.387175
},
{
"text": "two-in-one Chromebook",
"relevance": 0.385518
},
{
"text": "nightmare fuel",
"relevance": 0.385284
},
{
"text": "mouse-first OS—not",
"relevance": 0.385193
},
{
"text": "parallel tasks",
"relevance": 0.381923
},
{
"text": "budget device",
"relevance": 0.380932
},
{
"text": "iPad",
"relevance": 0.35313
},
{
"text": "news",
"relevance": 0.333007
},
{
"text": "strides",
"relevance": 0.319667
},
{
"text": "iOS",
"relevance": 0.318235
},
{
"text": "thanks",
"relevance": 0.316534
}
]
[
{
"type": "Company",
"text": "Google",
"relevance": 0.385564,
"disambiguation": {
"subtype": [
"AcademicInstitution",
"AwardPresentingOrganization",
"OperatingSystemDeveloper",
"ProgrammingLanguageDeveloper",
"SoftwareDeveloper",
"VentureFundedCompany"
],
"name": "Google",
"dbpedia_resource": "http://dbpedia.org/resource/Google"
},
"count": 9
},
{
"type": "Company",
"text": "Samsung",
"relevance": 0.204475,
"disambiguation": {
"subtype": [],
"name": "Samsung",
"dbpedia_resource": "http://dbpedia.org/resource/Samsung"
},
"count": 4
},
{
"type": "Location",
"text": "Chromebooks",
"relevance": 0.129986,
"disambiguation": {
"subtype": [
"City"
]
},
"count": 2
},
{
"type": "Company",
"text": "Amazon",
"relevance": 0.119948,
"disambiguation": {
"subtype": [],
"name": "Amazon.com",
"dbpedia_resource": "http://dbpedia.org/resource/Amazon.com"
},
"count": 2
},
{
"type": "Location",
"text": "US",
"relevance": 0.109124,
"disambiguation": {
"subtype": [
"Region",
"AdministrativeDivision",
"GovernmentalJurisdiction",
"FilmEditor",
"Country"
],
"name": "United States",
"dbpedia_resource": "http://dbpedia.org/resource/United_States"
},
"count": 1
},
{
"type": "Company",
"text": "Apple",
"relevance": 0.108271,
"disambiguation": {
"subtype": [
"Brand",
"OperatingSystemDeveloper",
"ProcessorManufacturer",
"ProgrammingLanguageDesigner",
"ProgrammingLanguageDeveloper",
"ProtocolProvider",
"SoftwareDeveloper",
"VentureFundedCompany",
"VideoGameDeveloper",
"VideoGamePublisher"
],
"name": "Apple Inc.",
"dbpedia_resource": "http://dbpedia.org/resource/Apple_Inc."
},
"count": 1
},
{
"type": "Quantity",
"text": "$500",
"relevance": 0.0746897,
"count": 1
},
{
"type": "Quantity",
"text": "$50",
"relevance": 0.0746897,
"count": 1
}
]
答案 0 :(得分：0)
这里有一些混乱。您没有特别好地处理Watson API响应，而且您似乎也误解了此响应的形状。
您的analyze函数处理来自Watson API调用的响应。 Watson API有助于将从服务器返回的JSON响应解析为Python对象，例如列表和dicts。但是，您的代码然后调用json.dumps将其转换回字符串。通过这种方式，您可以撤消Watson API为您完成的一些工作。不要在json.dumps上致电response，只需返回response即可。
（我猜你从官方的IBM文档here中得到了这个：它与json.dumps有相同的调用，但Python代码示例仅用于演示目的。）< / p>
这解释了您获得的错误：results因此是一个字符串，因此当您使用for item in results进行迭代时，每个item都是1个字符的字符串。
但是，进行此更改不足以使代码正常工作。接下来，我们必须查看发生错误的循环，因为它仍然存在问题：
for item in results:
keywords = item["keywords"]["text"]
entities = item["entities"]["text"]
此代码将results视为列表，列表中的每个item都是包含keywords属性和entities属性的字典。换句话说，它假设数据类似于以下内容：
[
{
"keywords": { "text": "abc123", ... },
"entities": { "text": "def456", ... },
...
},
{
"keywords": { "text": "cba321", ... },
"entities": { "text": "fed654", ... },
...
},
...
]
但是，如果您仔细查看the IBM documentation（与上面链接的页面相同的页面），响应将以不同的形式返回。看起来更像是这样：
{
"keywords": [
{ "text": "abc123", ... },
{ "text": "def456", ... }
],
"entities": [
{ "text": "ghi789", ... }
],
...
}
特别是，keywords和entities是顶级对象/字典下的单独列表，它们的长度可能不同。
而不是上面的循环，你可能想要更像以下内容：
for item in results["keywords"]:
keyword_text = item["text"]
for item in results["entities"]:
entity_text = item["text"]
但是，我不确定你的原始循环应该做什么：它会从你的Watson响应中获取数据，然后对它取出的数据不做任何事情。
您还需要修改Jinja模板以包含两个单独的循环。我会留给你的。
最后，您编写以下代码：
return render_template("results.html", results=results, URL=URL)
# check if URL is valid
if not results:
return apology("this is not a valid URL")
在我们返回后检查网址是否有效为时已晚！检查（if not results）无法访问，永远不会运行。将此检查移至行results = analyze(...)。