100% found this document useful (18 votes)
429 views

Download full Graph Powered Machine Learning 1st Edition Alessandro Negro ebook all chapters

Powered

Uploaded by

denciclozev
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
100% found this document useful (18 votes)
429 views

Download full Graph Powered Machine Learning 1st Edition Alessandro Negro ebook all chapters

Powered

Uploaded by

denciclozev
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 40

Get ebook downloads in full at ebookmeta.

com

Graph Powered Machine Learning 1st Edition


Alessandro Negro

https://ebookmeta.com/product/graph-powered-machine-
learning-1st-edition-alessandro-negro/

OR CLICK BUTTON

DOWNLOAD NOW

Explore and download more ebook at https://ebookmeta.com


Recommended digital products (PDF, EPUB, MOBI) that
you can download immediately if you are interested.

Graph-Powered Analytics and Machine Learning with


TigerGraph: Driving Business Outcomes with Connected Data
1st Edition Victor Lee
https://ebookmeta.com/product/graph-powered-analytics-and-machine-
learning-with-tigergraph-driving-business-outcomes-with-connected-
data-1st-edition-victor-lee/
ebookmeta.com

Knowledge Graphs Applied - MEAP V02 Alessandro Negro

https://ebookmeta.com/product/knowledge-graphs-applied-
meap-v02-alessandro-negro/

ebookmeta.com

Machine Learning Zhou

https://ebookmeta.com/product/machine-learning-zhou/

ebookmeta.com

An Inquiry into the Nature and Causes of the Wealth of


Nations 1776 1st Edition Adam Smith

https://ebookmeta.com/product/an-inquiry-into-the-nature-and-causes-
of-the-wealth-of-nations-1776-1st-edition-adam-smith-5/

ebookmeta.com
Emerging Adulthood and Higher Education A New Student
Development Paradigm 1st Edition Joseph L Murray Editor
Jeffrey Jensen Arnett Editor
https://ebookmeta.com/product/emerging-adulthood-and-higher-education-
a-new-student-development-paradigm-1st-edition-joseph-l-murray-editor-
jeffrey-jensen-arnett-editor/
ebookmeta.com

Stop Stress Improving the quality of your life A new


approach to stress and anxiety 1st Edition Marcel Verheyen

https://ebookmeta.com/product/stop-stress-improving-the-quality-of-
your-life-a-new-approach-to-stress-and-anxiety-1st-edition-marcel-
verheyen/
ebookmeta.com

The Psychology of Law Lubogo Christopher Isaac

https://ebookmeta.com/product/the-psychology-of-law-lubogo-
christopher-isaac/

ebookmeta.com

The Hidden History of Code Breaking 1st Edition Sinclair


Mckay

https://ebookmeta.com/product/the-hidden-history-of-code-breaking-1st-
edition-sinclair-mckay/

ebookmeta.com

Handbook of Spinal Cord Injuries and Related Disorders A


Guide to Evaluation and Management 1st Edition Hyun Yoon
Ko Sungchul Huh
https://ebookmeta.com/product/handbook-of-spinal-cord-injuries-and-
related-disorders-a-guide-to-evaluation-and-management-1st-edition-
hyun-yoon-ko-sungchul-huh/
ebookmeta.com
Women Talk Back to Shakespeare New Interdisciplinary
Approaches to Early Modern Culture 1st Edition Jo
Eldridge Carney
https://ebookmeta.com/product/women-talk-back-to-shakespeare-new-
interdisciplinary-approaches-to-early-modern-culture-1st-edition-jo-
eldridge-carney/
ebookmeta.com
Alessandro Negro
Foreword by Dr. Jim Webber

MANNING
Graph-Powered Machine Learning
Graph-Powered
Machine Learning
ALESSANDRO NEGRO
FOREWORD BY DR. JIM WEBBER

MANNING
SHELTER ISLAND
For online information and ordering of this and other Manning books, please visit
www.manning.com. The publisher offers discounts on this book when ordered in quantity.
For more information, please contact
Special Sales Department
Manning Publications Co.
20 Baldwin Road
PO Box 761
Shelter Island, NY 11964
Email: orders@manning.com

©2021 by Manning Publications Co. All rights reserved.

No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in


any form or by means electronic, mechanical, photocopying, or otherwise, without prior written
permission of the publisher.

Many of the designations used by manufacturers and sellers to distinguish their products are
claimed as trademarks. Where those designations appear in the book, and Manning
Publications was aware of a trademark claim, the designations have been printed in initial caps
or all caps.

Recognizing the importance of preserving what has been written, it is Manning’s policy to have
the books we publish printed on acid-free paper, and we exert our best efforts to that end.
Recognizing also our responsibility to conserve the resources of our planet, Manning books are
printed on paper that is at least 15 percent recycled and processed without the use of elemental
chlorine.

The author and publisher have made every effort to ensure that the information in this book
was correct at press time. The author and publisher do not assume and hereby disclaim any
liability to any party for any loss, damage, or disruption caused by errors or omissions, whether
such errors or omissions result from negligence, accident, or any other cause, or from an usage
of the information herein.

Manning Publications Co. Development editor: Dustin Archibald


20 Baldwin Road Technical development editors: Michiel Trimpe & Al Krinker
PO Box 761 Review editor: Ivan Martinović
Shelter Island, NY 11964 Production editor: Andy Marinkovich
Copy editor: Keir Simpson
Proofreader: Katie Tennant
Technical proofreader: Alex Ott
Typesetter: Gordan Salinovic
Cover designer: Marija Tudor

ISBN 9781617295645
Printed in the United States of America
To Filippo and Flavia:
I hope you are as proud of your father
as I am always proud of you.
brief contents
PART 1 INTRODUCTION ................................................................1
1 ■ Machine learning and graphs: An introduction 3
2 ■ Graph data engineering 30
3 ■ Graphs in machine learning applications 71

PART 2 RECOMMENDATIONS ......................................................113


4 ■ Content-based recommendations 119
5 ■ Collaborative filtering 166
6 ■ Session-based recommendations 202
7 ■ Context-aware and hybrid recommendations 227

PART 3 FIGHTING FRAUD ..........................................................263


8 ■ Basic approaches to graph-powered fraud detection 265
9 ■ Proximity-based algorithms 295
10 ■ Social network analysis against fraud 320

vii
viii BRIEF CONTENTS

PART 4 TAMING TEXT WITH GRAPHS ............................................357


11 ■ Graph-based natural language processing 359
12 ■ Knowledge graphs 389
contents
foreword xiii
preface xv
acknowledgments xvii
about this book xix
about the author xxiii
about the cover illustration xxiv

PART 1 INTRODUCTION ......................................................1

1 Machine learning and graphs: An introduction


1.1 Machine learning project life cycle 5
Business understanding 7 ■
Data understanding 8 Data
3

preparation 8 Modeling

9 Evaluation 9 Deployment
■ ■
9
1.2 Machine learning challenges 10
The source of truth 10 Performance ■
13 ■
Storing the
model 14 Real time 14

1.3 Graphs 15
What is a graph? 15 ■
Graphs as models of networks 17
1.4 The role of graphs in machine learning 23
Data management 25 ■
Data analysis 25 ■
Data
visualization 26
1.5 Book mental model 27

ix
x CONTENTS

2 Graph data engineering


2.1 Working with big data
Volume 34 ■
Velocity
30

36
33

Variety 38 ■
Veracity 39
2.2 Graphs in the big data platform 40
Graphs are valuable for big data 41 ■
Graphs are valuable for
master data management 48
2.3 Graph databases 53
Graph database management 54 Sharding 57 Replication 60
■ ■

Native vs. non-native graph databases 61 Label property graphs 67 ■

3 Graphs in machine learning applications


3.1 Graphs in the machine learning workflow 73
71

3.2 Managing data sources 76


Monitor a subject 79 Detect a fraud 82
■ ■
Identify risks in a
supply chain 85 Recommend items 87

3.3 Algorithms 93
Identify risks in a supply chain 93 Find keywords in a

document 96 Monitor a subject 98


3.4 Storing and accessing machine learning models 100


Recommend items 101 ■
Monitoring a subject 103
3.5 Visualization 106
3.6 Leftover: Deep learning and graph neural networks 109

PART 2 RECOMMENDATIONS ............................................113

4 Content-based recommendations
4.1 Representing item features 122
119

4.2 User modeling 136


4.3 Providing recommendations 143
4.4 Advantages of the graph approach 164

5 Collaborative filtering
5.1
166
Collaborative filtering recommendations 170
5.2 Creating the bipartite graph for the User-Item dataset 172
5.3 Computing the nearest neighbor network 177
5.4 Providing recommendations 189
CONTENTS xi

5.5 Dealing with the cold-start problem 194


5.6 Advantages of the graph approach 198

6 Session-based recommendations
6.1 The session-based approach
202
203
6.2 The events chain and the session graph 206
6.3 Providing recommendations 212
Item-based k-NN 213 ■
Session-based k-NN 219
6.4 Advantages of the graph approach 224

7 Context-aware and hybrid recommendations


7.1 The context-based approach 228
Representing contextual information 231 Providing ■
227

recommendations 235 Advantages of the graph approach



253
7.2 Hybrid recommendation engines 254
Multiple models, single graph 256 Providing ■

recommendations 258 Advantages of the graph approach



260

PART 3 FIGHTING FRAUD ................................................263

8 Basic approaches to graph-powered fraud detection


8.1 Fraud prevention and detection 267
265

8.2 The role of graphs in fighting fraud 271


8.3 Warm-up: Basic approaches 279
Finding the origin point of credit card fraud 279 Identifying a ■

fraud ring 287 Advantages of the graph approach 293


9 Proximity-based algorithms
9.1
295
Proximity-based algorithms: An introduction 296
9.2 Distance-based approach 298
Storing transactions as a graph 300 Creating the k-nearest

neighbors graph 302 Identifying fraudulent transactions 309


Advantages of the graph approach 318

10 Social network analysis against fraud


10.1 Social network analysis concepts 323
320

10.2 Score-based methods 326


Neighborhood metrics 330 Centrality metrics

336
Collective inference algorithms 344
xii CONTENTS

10.3 Cluster-based methods 348


10.4 Advantages of graphs 354

PART 4 TAMING TEXT WITH GRAPHS ..................................357

11 Graph-based natural language processing 359


11.1 A basic approach: Store and access sequence of words 363
Advantages of the graph approach 373
11.2 NLP and graphs 373
Advantages of the graph approach 387

12 Knowledge graphs
12.1
389
Knowledge graphs: Introduction 390
12.2 Knowledge graph building: Entities 393
12.3 Knowledge graph building: Relationships 402
12.4 Semantic networks 409
12.5 Unsupervised keyword extraction 415
Keyword co-occurrence graph 423 ■
Clustering keywords and topic
identification 425
12.6 Advantages of the graph approach 428

appendix A Machine learning algorithms taxonomy 431


appendix B Neo4j 435
appendix C Graphs for processing patterns and workflows 449
appendix D Representing graphs 458
index 461
foreword
The technology world is abuzz with machine learning. Every day we are bombarded
with articles on its applications and advances. But there is a quiet revolution brewing
among practitioners, and that revolution puts graphs at the very heart of machine
learning.
Alessandro wrote this book after almost a decade of practice, at the confluence of
graphs and machine learning. Had Alessandro worked for one of the Web giants dis-
tilling the knowledge of an army of PhDs working on special one-off systems, this
would be an interesting book, but for the majority of us it would satisfy our curiosity
rather than being a practical guide. Fortunately for us, while Alessandro does have a
PhD, he works in the enterprise space and has deep empathy and understanding for
the kinds of systems that enterprises build. The book reflects this: Alessandro ably
addresses the kinds of practical design and implementation challenges that software
engineers and data professionals building contemporary systems outside of the hyper-
scale Web giants must circumvent.
Graph-Powered Machine Learning demonstrates how important graphs are to the
future of machine learning. It shows not only that graphs provide a superior means of
fuelling contemporary ML pipelines, but also how graphs are a natural way of organiz-
ing, analyzing, and processing data for machine learning. The book offers a rich,
curated tour of graph machine learning, and each topic is underpinned with detailed
examples drawing on Alessandro’s deep experience and the easy, refined confidence
of a long-term practitioner.

xiii
xiv FOREWORD

The book eases us in, providing an overall framework to reason about machine
learning and integrate it into our data systems. It follows up immediately with a practi-
cal approach to recommendations covering a variety of approaches, such as collabora-
tive filtering, content- and session-based recommendations, and hybrid styles.
Alessandro calls out the problems which lack explainability in state-of-the-art tech-
niques and shows that this isn’t an issue with the graph approach. He then continues
to tackle fraud detection, taking in concepts like proximity and social network analy-
sis, where we relearn the maxim that “birds of a feather flock together” in the context
of criminal networks. Finally, the book deals with knowledge graphs: the ability of
graph technology to consume documents and distil connected knowledge from them,
disambiguate terms, and handle ambiguous query terms. The breadth of topics is vast,
but the quality of information is always excellent.
Throughout the book, Alessandro gently guides the reader, building up from the
basics to advanced concepts. With the examples and companion code, practically
minded readers are able to get examples working quickly, and from there to adapt
them for their own needs. You will finish this book armed with a variety of practical
tools at your disposal and, if you like, some dirt under your fingernails. You will be
ready to extract graph features to make your existing models perform better today,
and you’ll be equipped to work natively with graphs tomorrow. I promise it’s going to
be a wonderful journey.
—DR. JIM WEBBER, CHIEF SCIENTIST @ NEO4J
Other documents randomly have
different content
PLATE XCV

GREAT LOBELIA.—L. syphilitica.

The genus is named after an early Flemish herbalist, de l’Obel.

Indian Tobacco.
Lobelia inflata. Lobelia Family.

One to two feet high. Stem.—Branching from the root. Leaves.—Alternate,


ovate, or oblong, somewhat toothed. Flowers.—Blue or purple, small, growing in a
loose raceme, resembling in structure those of the great lobelia. Pod.—Much
inflated.
During the summer we note in the dry, open fields the blue
racemes of the Indian tobacco, and in the later year the inflated pods
which give it its specific name. The plant is said to be poisonous if
taken internally, and yields a “quack-medicine” of some notoriety.
The Indians smoked its dried leaves, which impart to the tongue a
peculiar tobacco-like sensation.
There are other species of lobelia which may be distinguished by
their narrower leaves and uninflated pods, and by their choice of
moist localities.

Hog Pea-nut.
Amphicarpæa monoica. Pulse Family (p. 16).

Stem.—Climbing and twining over plants and shrubs. Leaves.—Divided into


three somewhat four-sided leaflets. Flowers.—Papilionaceous, pale lilac, or
purplish, in nodding racemes. Pod.—One inch long.
Along the shadowy lanes which wind through the woods the
climbing members of the Pulse family are very abundant. During the
late summer and autumn the lonely wayside is skirted by
Vines, with clust’ring bunches growing;
Plants, with goodly burden bowing.

And in and out among this luxuriant growth twist the slender stems
of the ill-named hog pea-nut, its delicate lilac blossoms nodding
from the coarse stalks of the golden-rods and iron-weeds or blending
with the purple asters.
This plant bears flowers of two kinds: the upper ones are perfect,
but apparently useless, as they seldom ripen fruit; while the lower or
subterranean ones are without petals or attractiveness of
appearance, but yield eventually at least one large ripe seed.
PLATE XCVI

INDIAN TOBACCO.—L. inflata.

Beach Pea.
Lathyrus maritimus. Pulse Family (p. 16).

About one foot high, or more. Stem.—Stout. Leaves.—Divided into from three
to five pairs of thick oblong leaflets. Flowers.—Papilionaceous, large, purple,
clustered.
The deep-hued flowers of this stout plant are commonly found
along the sand-hills of the seashore, and also on the shores of the
Great Lakes, blooming in early summer. Both flowers and leaves are
at once recognized as belonging to the Pulse family.
——— ———
Strophostyles angulosa. Pulse Family (p. 16).

Stems.—Branched, one to six feet long, prostrate or climbing. Leaves.—


Divided into three leaflets, which are more or less prominently lobed toward the
base, the terminal two-lobed; or some or all without lobes. Flowers.—Purplish or
greenish, on long flower-stalks. Pod.—Linear, straight, or nearly so.
This somewhat inconspicuous plant is found back of the sand-
hills along the coast, often in the neighborhood of the beach pea, and
climbing over river-banks, thickets, and fences as well. It can usually
be identified by its oddly lobed leaflets.

Blue Vetch.
Vicia cracca. Pulse Family (p. 16).

Leaves.—Divided into twenty to twenty-four leaflets, with slender tips.


Flowers.—Papilionaceous, blue turning purple, growing in close, many-flowered,
one-sided spikes.
This is an emigrant from Europe which is found in some of our
eastern fields and thickets as far south as New Jersey. It usually
climbs more or less by means of the tendril at the tip of its divided
leaves, and sometimes forms bright patches of vivid blue over the
meadows.
Another member of this genus is V. sativa, the common vetch or
tare, with purplish or pinkish flowers, growing singly or in pairs from
the axils of the leaves, which leaves are divided into fewer and
narrower leaflets than those of the blue vetch. This species also takes
possession of cultivated fields as well as of waste places along the
roadside.
PLATE XCVII

BEACH PEA.—L. maritimus.

Chicory. Succory.
Cichorium Intybus. Composite Family (p. 13).

Stems.—Branching. Leaves.—The lower oblong or lance-shaped, partly


clasping, sometimes sharply incised, the floral ones minute. Flower-heads.—Blue,
set close to the stem, composed entirely of strap-shaped flowers; opening at
different times.

Oh, not in Ladies’ gardens,


My peasant posy!
Smile thy dear blue eyes,
Nor only—nearer to the skies—
In upland pastures, dim and sweet,—
But by the dusty road
Where tired feet
Toil to and fro;
Where flaunting Sin
May see thy heavenly hue,
Or weary Sorrow look from thee
Toward a more tender blue![12]

This roadside weed blossoms in late summer. It is extensively


cultivated in France, where the leaves are blanched and used in a
salad which is called “Barbe des Capucins.” The roots are roasted and
mixed with coffee both there and in England.
Horace mentions its leaves as part of his frugal fare, and Pliny
remarks upon the importance of the plant to the Egyptians, who
formerly used it in great quantities, and of whose diet it is still a
staple article.

Blue and Purple Asters.


Aster. Composite Family (p. 13).

Flower-heads.—Composed of blue or purple ray-flowers, with a centre of


yellow disk-flowers.
PLATE XCVIII

CHICORY.—C. Intybus.

As about one hundred and twenty different species of aster are


native to the United States, and as fifty-four of these are found in
Northeastern America, all but a dozen being purple or blue (i.e., with
purple or blue ray-flowers), and as even botanists find that it
requires patient application to distinguish these many species, only a
brief description of the more conspicuous and common ones is here
attempted.
Along the dry roadsides in early August we may look for the
bright blue-purple flowers of A. patens. This is a low-growing
species, with rough, narrowly oblong, clasping leaves, and widely
spreading branches, whose slender branchlets are usually terminated
by a solitary flower-head.
Probably no member of the group is more striking than the New
England aster, A. Novæ Angliæ, whose stout hairy stem (sometimes
eight feet high), numerous lance-shaped leaves, and large violet-
purple or sometimes pinkish flower-heads, are conspicuous in the
swamps of late summer.
A. puniceus is another tall swamp-species, with long showy pale
lavender ray-flowers.
One of the most commonly encountered asters is A. cordifolius,
which is far from being the only heart-leaved species, despite its title.
Its many small, pale blue or almost white flower-heads mass
themselves abundantly along the wood-borders and shaded
roadsides.
Perhaps the loveliest of all the tribe is the seaside purple aster,
A. spectabilis, a low plant with narrowly oblong leaves and large
bright heads, the violet-purple ray-flowers of which are nearly an
inch long. This grows in sandy soil near the coast and may be found
putting forth its royal, daisy-like blossoms into November.
Great Britain can claim but one native aster, A. Trifolium, or
sea-starwort as it is called. Many American species are cultivated in
English gardens under the general title of Michaelmas daisies. The
starwort of Italy is A. amellus. The Swiss species is A. Alpinum.
This beautiful genus, like that of the golden-rod, is one of the
peculiar glories of our country. Every autumn these two kinds of
flowers clothe our roadsides and meadows with so regal a mantle of
purple and gold that we cannot but wonder if the flowers of any other
region combine in such a radiant display.

Iron-weed.
Vernonia Noveboracensis. Composite Family (p. 13).

Stem.—Leafy, usually tall. Leaves.—Alternate, somewhat lance-oblong.


Flower-heads.—An intense red-purple, loosely clustered, composed entirely of
tubular flowers.
Along the roadsides and low meadows near the coast the iron-
weed adds its deep purple hues to the color-pageant of late August.
By the uninitiated the plant is often mistaken for an aster, but a
moment’s inspection will discover that the minute flowers which
compose each flower-head are all tubular in shape, and that the ray
or strap-shaped blossoms which an aster must have are wanting.
These flower-heads are surrounded by an involucre composed of
small scales which are tipped with a tiny point and are usually of a
purplish color also.

Blue Curls. Bastard Pennyroyal.


Trichostema dichotomum. Mint Family (p. 16).

Stem.—Rather low, branching, clammy. Leaves.—Opposite, narrowly oblong


or lance-shaped, glutinous, with a balsamic odor. Flowers.—Purple, occasionally
pinkish, not usually clustered. Calyx.—Five-cleft, two-lipped. Corolla.—Five-lobed,
the three lower lobes more or less united. Stamens.—Four, very long and curved,
protruding. Pistil.—One, with a two-lobed style.
In the sandy fields of late summer this little plant attracts notice
by its many purple flowers. Its corolla soon falls and exposes to view
the four little nutlets of the ovary lying within the enlarged calyx like
tiny eggs in their nest. Its aromatic odor is very perceptible, and the
little glands with which it is covered may be seen with the aid of a
magnifier. The generic name, Trichostema, signifies hairy stamens
and alludes to the curved hair-like filaments.

Sea Lavender. Marsh Rosemary.


Statice Caroliniana. Leadwort Family.

Stems.—Leafless, branching. Leaves.—From the root, somewhat oblong, thick.


Flowers.—Lavender-color or pale purple, tiny, scattered or loosely spiked along
one side of the branches. Calyx.—Dry, funnel-form. Corolla.—Small, with five
petals. Stamens.—Five. Pistil.—One, with five, rarely three, styles.
In August many of the salt marshes are blue with the tiny
flowers of the sea lavender. The spray-like appearance of the little
plant would seem to account for its name of rosemary, which is
derived from the Latin for sea-spray, but Dr. Prior states that this
name was given it on account of “its usually growing on the sea-
coast, and its odor.”
Blossoming with the lavender we often find the great rose
mallows and the dainty sea pinks. The marsh St. John’s-wort as well
is frequently a neighbor, and, a little later in the season, the salt
marsh fleabane.

Blazing Star.
Liatris scariosa. Composite Family (p. 13).

Stem.—Simple, stout, hoary, two to five feet high. Leaves.—Alternate,


narrowly lance-shaped. Flower-heads.—Racemed along the upper part of the stem,
composed entirely of tubular flowers of a beautiful shade of rose-purple.
These showy and beautiful flowers lend still another tint to the
many-hued salt marshes and glowing inland meadows of the falling
year. Gray assigns them to dry localities from New England to
Minnesota and southward, while my own experience of them is
limited to the New England coast, where their stout leafy stems and
bright-hued blossoms are noticeable among the golden-rods and
asters of September. The hasty observer sometimes confuses the
plant with the iron-weed, but the two flowers are very different in
color and in their manner of growth.

Common Dittany.
Cunila Mariana. Mint Family (p. 16).

About one foot high. Stem.—Much branched, reddish. Leaves.—Opposite,


aromatic, dotted, smooth, ovate, rounded or heart-shaped at base, set close to the
stem. Flowers.—Small, purple, lilac or white, clustered. Calyx.—Five-toothed.
Corolla.—Small, two-lipped, the upper lip erect, usually notched, the lower three-
cleft. Stamens.—Two, erect, protruding. Pistil.—One, with a two-lobed style.
In late August or early September the delicate flowers of the
dittany brighten the dry, sterile banks which flank so many of our
roadsides. At a season when few plants are flowering save the
omnipresent members of the great Composite family these dainty
though unpretentious blossoms are especially attractive. The plant
has a pleasant fragrance.
PLATE XCIX

BLAZING STAR.—L. scariosa.

Closed Gentian.
Gentiana Andrewsii. Gentian Family.

Stem.—One to two feet high, upright, smooth. Leaves.—Opposite, narrowly


oval or lance-shaped. Flowers.—Blue to purple, clustered at the summit of the
stem and often in the axils of the leaves. Calyx.—Four or five-cleft. Corolla.—
Closed at the mouth, large, oblong. Stamens.—Four or five. Pistil.—One, with two
stigmas.
Few flowers adapt themselves better to the season than the
closed gentian. We look for it in September when the early waning
days and frost-suggestive nights prove so discouraging to the greater
part of the floral world. Then in somewhat moist, shaded places
along the roadside we find this vigorous, autumnal-looking plant,
with stout stems, leaves that bronze as the days advance, and deep-
tinted flowers firmly closed as though to protect the delicate
reproductive organs within from the sharp touches of the late year.
To me the closed gentian usually shows a deep blue or even
purple countenance, although like the fringed gentian and so many
other flowers its color is lighter in the shade than in the sunlight. But
Thoreau claims for it a “transcendent blue,” “a splendid blue, light in
the shade, turning to purple with age.” “Bluer than the bluest sky,
they lurk in the moist and shady recesses of the banks,” he writes.
Mr. Burroughs also finds it “intensely blue.”

Five-flowered Gentian.
Gentiana quinqueflora. Gentian Family.

Stem.—Slender, branching, one or two feet high. Leaves.—Opposite, ovate,


lance-shaped, partly clasping. Flowers.—Pale blue, smaller than those of the closed
gentian, in clusters of about five at the summit of stems and branches. Calyx.—
Four or five-cleft, small. Corolla.—Funnel-form, four or five-lobed, its lobes
bristle-pointed. Stamens.—Four or five. Pistil.—One, with two stigmas.
Although the five-flowered gentian is far less frequently
encountered than the closed gentian, it is very common in certain
localities. Gray assigns it to “moist hills” and “along the mountains to
Florida.” I have found it growing in great abundance on the
Shawangunk Mountains in Orange County, N. Y., where it flowers in
September.
PLATE C

CLOSED GENTIAN.—G. Andrewsii.

Fringed Gentian.
Gentiana crinita. Gentian Family.

Stem.—One to two feet high. Leaves.—Opposite, lance-shaped or narrowly


oval. Flowers.—Blue, large. Calyx.—Four-cleft, the lobes unequal. Corolla.—
Funnel-form, with four fringed, spreading lobes. Stamens.—Four. Pistil.—One,
with two stigmas.
In late September when we have almost ceased to hope for new
flowers we are in luck if we chance upon this
—blossom bright with autumn dew
whose
—sweet and quiet eye
Looks through its fringes to the sky,
Blue—blue—as if that sky let fall,
A flower from its cerulean wall;[13]

for the fringed gentian is fickle in its habits, and the fact that we have
located it one season does not mean that we will find it in the same
place the following year; being a biennial, with seeds that are easily
washed away, it is apt to change its haunts from time to time. So our
search for this plant is always attended with the charm of
uncertainty. Once having ferreted out its new abiding-place,
however, we can satiate ourselves with its loveliness, which it usually
lavishes unstintingly upon the moist meadows which it has elected to
honor.
Thoreau describes its color as “such a dark blue! surpassing that
of the male bluebird’s back!” My experience has been that the flowers
which grow in the shade are of a clear pure azure, “Heaven’s own
blue,” as Bryant claims; while those which are found in open, sunny
meadows may be justly said to vie with the back of the male bluebird.
If the season has been a mild one we shall perhaps find a few
blossoms lingering into November, but the plant is probably blighted
by a severe frost, although Miss Emily Dickinson’s little poem voices
another opinion:
· · · · ·

But just before the snows


There came a purple creature
That ravished all the hill:
And Summer hid her forehead,
And mockery was still.
The frosts were her condition:
The Tyrian would not come
Until the North evoked it,
“Creator! shall I bloom?”
PLATE CI

FRINGED GENTIAN.—G. crinita.


VI
MISCELLANEOUS

Skunk Cabbage. Swamp Cabbage.


Symplocarpus fœtidus. Arum Family.

Leaves.—Large, becoming one or two feet long; heart-shaped, appearing later


than the purple-mottled spathe and hidden flowers. Flowers.—Small and
inconspicuous; packed on the fleshy spike which is hidden within the spathe.
If we are bold enough to venture into certain swampy places in
the leafless woods and brown cheerless meadows of March, we notice
that the sharply pointed spathes of the skunk cabbage have already
pierced the surface of the earth. Until I chanced upon a passage in
Thoreau’s Journal under date of October 31st, I had supposed that
these “hermits of the bog” were only encouraged to make their
appearance by the advent of those first balmy, spring-suggestive days
which occasionally occur as early as February. But it seems that
many of these young buds had pushed their way upward before the
winter set in, for Thoreau counsels those who are afflicted with the
melancholy of autumn to go to the swamps, “and see the brave
spears of skunk-cabbage buds already advanced toward a new year.”
“Mortal and human creatures must take a little respite in this fall of
the year,” he writes. “Their spirits do flag a little. There is a little
questioning of destiny, and thinking to go like cowards to where the
weary shall be at rest. But not so with the skunk-cabbage. Its
withered leaves fall and are transfixed by a rising bud. Winter and
death are ignored. The circle of life is complete. Are these false
prophets? Is it a lie or a vain boast underneath the skunk-cabbage
bud pushing it upward and lifting the dead leaves with it?”
PLATE CII

SKUNK CABBAGE.—S. fœtidus.

The purplish shell-like leaf, which curls about the tiny flowers
which are thus hidden from view, is a rather grewsome-looking
object, suggestive of a great snail when it lifts itself fairly above its
muddy bed. When one sees it grouped with brother-cabbages it is
easy to understand why a nearly allied species, which abounds along
the Italian Riviera, should be entitled “Cappucini” by the neighboring
peasants, for the bowed, hooded appearance of these plants might
easily suggest the cowled Capuchins.
It seems unfortunate that our earliest spring flower (for such it
undoubtedly is) should possess so unpleasant an odor as to win for
itself the unpoetic title of skunk cabbage. There is also some
incongruity in the heading of the great floral procession of the year
by the minute hidden blossoms of this plant. That they are enabled to
survive the raw March winds which are rampant when they first
appear is probably due to the protection afforded them by the
leathery leaf or spathe. When the true leaves unfold they mark the
wet woods and meadows with bright patches of rich foliage, which
with that of the hellebore, flash constantly into sight as we travel
through the country in April.
It is interesting to remember that the skunk cabbage is nearly
akin to the spotless calla lily, the purple-mottled spathe of the one
answering to the snowy petal-like leaf of the other. Meehan tells us
that the name bear-weed was given to the plant by the early Swedish
settlers in the neighborhood of Philadelphia. It seems that the bears
greatly relished this early green, which Meehan remarks “must have
been a hot morsel, as the juice is acrid, and is said to possess some
narcotic power, while that of the root, when chewed, causes the
eyesight to grow dim.”

Wild Ginger.
Asarum Canadense. Birthwort Family.

Leaves.—One or two on each plant, kidney or heart-shaped, fuzzy, long-


stalked. Flower.—Dull purplish-brown, solitary, close to the ground on a short
flower-stalk from the fork of the leaves. Calyx.—Three-cleft, bell-shaped. Corolla.
—None. Stamens.—Twelve. Pistil.—One, with a thick style and six thick, radiating
stigmas.
PLATE CIII

WILD GINGER.—A. Canadense.

Certain flowers might be grouped under the head of “vegetable


cranks.” Here would be classed the evening primrose, which only
opens at night, the closed gentian, which never opens at all, and the
wild ginger, whose odd, unlovely flower seeks protection beneath its
long-stemmed fuzzy leaves, and hides its head upon the ground as if
unwilling to challenge comparison with its more brilliant brethren.
Unless already familiar with this plant there is nothing to tell one
when it has reached its flowering season; and many a wanderer
through the rocky woods in early May quite overlooks its shy,
shamefaced blossom.
The ginger-like flavor of the rootstock is responsible for its
common name. It grows wild in many parts of Europe and is
cultivated in England, where at one time it was considered a remedy
for headache and deafness.

You might also like